Book of Proceedings - UBT

Transcription

Proceedings of
4 UBT Annual International Conference on
Business, Technology and Innovation
th
Chapter: Computer Science and Communication Engineering
Chapter: Information Systems and Security
November, 2015
1
ISBN 978-9951-550-14-7
© UBT – Higher Education Institution
International Conference on Business, Technology and
Innovation Durres, Albania 6-7 November 2015
Editor: Edmond Hajrizi
Organizers: Albert Qarri, Felix Breitenecker, Krenare Pireva, Evelina Bazini,
Kozeta Sevrani, Betim Gashi
Co-organizes: Ardian Emini, Muzafer Shala, Lulzim Beqiri, Mimoza
Sylejmani, Besian Sinani, Xhemajl Mehmeti, Murat
Retkoceri, Bertan Karahoda, Ermal Lubishtani, Albulena
Jahja, Erveina Gosalci, Alfred Marleku, Ibrahim Krasniqi,
Ylber Limani, Naim Preniqi, Rexhep Kera, Muhamet
Sherifi, Ermal Gashi
Authors themselves are responsible for the integrity of what is being published.
Copyright © 2015 UBT. All rights reserve
Publisher, UBT
2
EDITOR SPEECH
International Conference on Business, Technology and Innovation is an
international interdisciplinary peer reviewed conference which publishes
works of the scientists as well as practitioners in different research area.
The main perspective of the conference is to connect the scientists and
practitioners from different disciplines in the same place and make them
be aware of the recent advancements in different research fields, and
provide them with a unique forum to share their experiences.
It is also the place to support the new academic staff for doing research and
publish their work in international standard level.
This conference consists of 10 sub conferences in different fields:










Computer Science and Communication Engineering
Energy Efficiency Engineering
Management, Business and Economics
Mechatronics, System Engineering and Robotics
Information Systems and Security
Architecture - Spatial Planning and Civil Engineering
Civil Engineering, Infrastructure and Environment
Law
Political Science & International Relations
Journalism, Media and Communication
This conference is the major scientific event of the University for Business
and Technology.
It is organizing annually and always in cooperation with the partner
universities from the region and Europe. In this case our partner
universities are the University of Vlora “Ismail Qemaili” and University of
Tirana –Faculty of Economics. Other professional partners in this
3
conference are: EUROSIM, Kosova Association for Control, Automation and
Systems Engineering (KA – CASE), Kosova Association for Modeling and
Simulation (KA – SIM), Quality Kosova, Kosova Association for
Management.
We have to thank all Authors, partners, sponsors and also the conference
organizing team making this event a real international scientific event.
This year we have more participants and publication than last year. We
have paper from 15 different countries worldwide.
Congratulations!
November 2015
Prof. Dr.
Edmond Hajrizi
Chairman of ICBTI 2015, UBT
4
CONTENTS
IC-CSCE...................................................................................................................................... 8
Digital long-term preservation in Albania, determining the right file formats ................... 9
Ardiana Topi, Aleksandër Xhuvani
From Traditional Markets to E-Commerce and Finally to Social Media Commerce ....... 15
Ardian Hyseni
New integral transform in Caputo type fractional difference operator ............................. 23
Artion Kashuri, Akli Fundo, Rozana Liko
The mobile telecom operator T-Mobile Macedonia AD Skopje abuses its dominant position
harmful for the mobile telecom operators ONE and VIP in the telecom market .............. 29
Boris Arsov
The Analytic Hierarchic Process: A Model to Evaluate the Risk of Getting Cancer ........ 36
Daniela Qendraj (Halidini), Markela Muca
Implementation of E-education in Africa via Space Networks ........................................ 42
Dimov Stojce Ilcev
Introduction to Stratospheric Communication Platforms (SCP) ....................................... 50
Dimov Stojce Ilcev
The Importance of Big Data Analytics ............................................................................. 59
Eljona Proko
Assessing Clustering in a Social University Network Course .......................................... 64
Orgeta Gjermëni
Perfect Metamaterial absorber based energy harvesting application in ISM Band ........... 68
Furkan Dincer, Mehmet Bakir, Muharrem Karaaslan, Kemal Delihacioglu, Cumali Sabah
Web Application for Students and Lecturer of Electrical Engineering Faculty ................ 74
Genci Sharko, Anni Dasho Sharko, Kadri Manaj
Exploring the role of sentiments in identification of active and influential bloggers ....... 81
Mohammad Alghobiri, Umer Ishfaq, Hikmat Ullah Khan, Tahir Afzal Malik
Morphological parsing of Albanian language: a different approach to Albanian verbs ... 87
Ilir Çollaku, Eşref Adalı
A Policeless traffic ticketing system with autonomous vehicles ..................................... 92
Mükremin Özkul and Ilir Çapuni ....................................................................................................... 92
Labeled-Image CAPTCHA: concept of a secured and universally useful CAPTCHA .. 102
Mokter Hossain1, Ken Nguyen2, Muhammad Asadur Rahman2 ...................................................... 102
Dual Tone Multi Frequency (DTMF) signal generation and detection using MATLAB
software .......................................................................................................................... 109
Nihat Pamuk, Ziynet Pamuk
Randomizing Ensemble-based approaches for Outlier ................................................... 115
Lediona Nishani , Marenglen Biba
10 W DC-DC converter based LED driver circuit design .............................................. 120
Ö. Faruk Farsakoğlu, H. Yusuf Hasirci, İbrahim Çelik
The influence of cloud computing and mobile technology in our activities ................... 125
Ramadan Cikaqi, Ilir Morina
Bounty techniques for web vulnerability scanning......................................................... 131
Tanzer Abazi, Mentor Hoxhaj, Edmond Hajrizi, Gazmend Krasniqi
Security Concerns of new alternative telecommunication services ............................... 135
Arbnora Hyseni, Krenare Pireva, Miranda Kajtazi
5
International Conference on Computer Science and Communication Engineering, Nov 2015
Implementation of the AODV Routing in an Energy-constrained Mesh Network ......... 140
Altin Zejnullahu, Zhilbert Tafa
Exploring the role of sentiments in identification of active and influential bloggers ..... 146
Mohammad Alghobiri, Umer Ishfaq, Hikmat Ullah Khan, Tahir Afzal Malik
Internet of Things: From applications, challenges and standardization to Industry
implementations ............................................................................................................. 152
Xhafer Krasniqi
Case Studies for a Markov Chain Approach to Analyze Agent-Based Models .............. 166
Florian Kitzler, Martin Bicher
Hybrid Modelling in System Simulation ........................................................................ 171
Andreas Körner
Definition, Validation and Comparison of Two Population Models for Austria ............ 174
Martin Bicher, Barbara Glock, Florian Miksch, Niki Popper, Günter Schneckenreither
Microsimulation Models for Simulating Pathways of Patients with Mental Diseases ... 179
Andreas Bauer, Felix Breitenecker, Christoph Urach
IC-ISS ...................................................................................................................................... 183
E-learning systems in higher education institutions: An outlook of their use in the Western
Balkan Region ................................................................................................................ 184
Blerta Abazi Chaushi, Agron Chaushi, Florije Ismaili
Credit Information System in Albania............................................................................ 190
Valbona Çinaj, Bashkim Ruseti
Performance Indicators Analysis inside a Call Center Using a Simulation Program ..... 196
Ditila Ekmekçiu, Markela Muça, Adrian Naço
IT Outsourcing ............................................................................................................... 210
Besnik Skenderi, Diamanta Skenderi
Modelling business and management systems using Fuzzy cognitive maps: A critical
overview ......................................................................................................................... 213
Peter P. Groumpos
E-customer relationship management in insurance industry in Albania ......................... 221
Evelina Bazini
Some Principles for Banks’ Internal Control System in Albania ................................... 226
Artur Ribaj
6
International Conference on
Computer Science and
Communication Engineering
7
IC-CSCE
International Program Committee:





















E.Hajrizi (RKS), Chair
Baltes J.(CAN)
Breitenecker F.(AUT)
Gashi I.(UK)
Retschitzegger W. (AUT)
Hofbauer P.(AUT)
Jesse N.(GER)
Kopacek P.(AUT)
Krasniqi Xh. (UK)
Mili F. (USA)
Mirijamdotter A. (SWE)
Wagner C. (USA)
Schwaerzel H.(GER)
Seiler W. (AUT )
Sev rani K .(AL)
Stapelton L.(IRL)
Yayilgan S .Y.(NOR)
Yusuf S.(UK)
Tafa Zh.(MNE)
Shala M. (RKS)
Koerner A. (AUT)
National Organizing Committee:







E.Hajrizi (KOS), Chair
Pireva K. (KOS)
Bazini E. (AL)
Karahoda B.(RKS)
Hyso A . (AL)
Gashi B. (RKS)
Sherifi M. (RKS)
Editors:
Edmond Hajrizi (UBT) & Felix Breitenecker (AUT)
8
Digital long-term preservation in Albania, determining the
right file formats
Ardiana Topi1, 2, Aleksandër Xhuvani3
1
University of Tirana, Faculty of Natural Sciences, Department of Informatics,
Blvd. Zogu 1, Tirana, Albania
2
National Archives of Albania, Street Jordan Misja, 8303, Tirana, Albania
3
Polytechnic University of Tirana, Faculty of Information Technology,
Department of Computer Engineering, Nënë Tereza Square, Tirana, Albania
ardianatopi@yahoo.com1
Abstract. The National Archive of Albania (NAA) is the main institution in Albania, according to
legislation, responsible to create the rules for document life management produced by any
organization, and preserve them. Almost all archival documents stored in archival institutions, must
be saved for unlimited period of time, or forever. Preservation of digital documents is different from
preservation of paper documents. It is a very complex procedure directly connected with the
inevitable phenomenon of hardware and software obsolesces. A successful mission of the NAA must
have a well-defined strategy regarding the Digital Long-term Preservation. This paper aims to present
the possible file formats, able to use during the digitization process of paper based documents and
photography.
Technical characteristics of same useful file formats are analyzed. The study tells that sustainability
and quality factors are very important in digital file format selection for long term preservation.
Keywords: file format, image quality, digital collection
1. Introduction
When we speak for the digital preservation, the Archival Institutions must ensure not only the
preservation of their archival files but ensure that these files will be authentic and readable for long
time. This task is not simple in the digital world. Long-term Digital Preservation refers to the series
of managed activities necessary to ensure a continuous access to digital materials or at least to the
information contained in them for as long as necessary, in the most of cases indefinitely [1]. The
machine and software dependency of digital collections, the rapid changes in technology, the fragility
of the media, the ease manipulation makes the digital long term preservation a very complex task.
File formats are very important in the ability for reusing and accessing data into the future that’s
because we must ensure to produce to high quality digital files. Those files must be produced as the
best copy by a digitizing process, where “best” is defined with meeting to the objectives of a particular
project or program [2].
The file formats in this study are chosen based on useful file formats in the Institution of General
Directory of Albania [3]. The digital file formats vary from simple to complex format structures. As
the complexity of the individual formats increases, the requirements for the structure and
completeness of the documents must be adjusted accordingly and realized consistently [4].
Our study aims to give recommendations regarding the file formats that can be used for digital long
term preservation and suggest the evaluation methods that can help the establishment of national
level standards and guideline in digital preservation field.
9
2. The issue of digital preservation
Nowadays, there is a rapidly increasing volume of documents and information created in digital form.
In every office we are going to create the documents digitally. In the same time there are lots of
public institutions that have started digitized projects that are going to produce large amounts of
digital content. But how does the creation of a digital document works? To create a digital document
we need hardware devices (a computer and/or scanner) and an application that define the structure
and encoding of that document. Reuse or editing the same digital document, needs an application
that can distinguish the way in which this document is encoded and based in that information the
application will be able to render it to the monitor. In the most of cases each application generates its
own file format. Actually, a large number of file formats are used worldwide. The issue stands,
whether we are able to read those documents after a long time without missing information and how
to ensure that? The file format that is defined as “a standard way that information is encoded for
storage in a computer file” [5].
Since we can use different file format for the same types of documents the issue raised is which is
the best file format for long term preservation and how can we evaluate it?
To answer this question, initially we identified the types of documents preserved in archival
institutions1. The documents found in different forms are grouped in four different categories: paper
based, photography, audio and video tape files. To convert those types on digital version different
types of file format need to be used. Each of those formats is completely different in structure, in
encode/decode algorithms and offer different advantages or disadvantages.
In this paper are presented digital file formats that can be used for preservation of digitally converted
paper based documents and photography. The study was based on research in scientific literature
review, International Standards and Best Practices applied and results of our project. The tests are
accomplished on the factors that have consequences in image quality and storage capacity. Finally
we have collected information about sustainability factors [6] for every file format studied.
3. The implemented methodology
Three main issues were studied: the storage capacity, the quality and sustainability factors of file
formats. Those issues were studied based on random sampling of documents from NAA’s collections
and were grouped in three different categories: machine write documents, manuscripts and
photography sets.
We used a workstation connected with a flatbed scanner. The software used for capture and edit was
the most difficult task, due to the lack of definition of the “image quality”. The image itself can be
subjective (measured by human eyes) or objective (measured by calculate digital image properties)
[11]. In order to achieve an acceptable image quality, it is necessary examination of the whole image
processing chain especially the issues concerning the calibration of scanning and rendering
equipment (monito and printer). The quality is affected by many factors such are source material,
devices used to capture, the subsequent processing, and compression method A quality rating system
must be linked with defects appears on digitization and processing of digital information based on
evaluation on-screen comparing with originals. The technical characteristics that influence on the
image quality are: Resolution, Color space, Bid depth.
The resolution was selected based on Quality Index formula (QI = 0.013 dpi x h) introduced by
Kenney and Chapman (1995). This resolution benchmarking method has its roots in micrographics,
where standards for predicting image quality are based on the quality index too. QI provides a tool
for relating system resolution and legibility, where h is the weight of the smallest letter in document
(usually e). This method employs three levels: high level QI = 8; medium level QI =5; and marginal
1
The study is done on documents collections of ANA and here we are speaking for traditional
documents not digital born documents. It will be object of another study.
10
level QI = 3.6. For non-text material we adapted this formula by replacing the high of smallest
character in text with the smallest complete part that is considered essential to an understanding of
the entire page, for example width of the lines used to render and it can be different for a photography
or for a papyrus manuscript.
The storage capacity was evaluated based on the information collected after finishing the digitized
process for each case. The data are organized and illustrated in following tables. The sustainability
factors are evaluated based on information available for each in terms of different convention scoring.
One example is demonstrated in the paper.
4. Technical aspects of digital collections
The paper based documents in archives are found in different forms such as: textual documents
including manuscripts, artworks and photographic formats and are found in both black and white;
and color mode. For this category of documents the conversion from analog to digital form is
accomplished by using scanners or digital camera. We have chosen the scanners since the light
conditions are very crucial for digital camera and have direct influence on image quality. The devices
in scanning system (monitors and scanners) are calibrated using ISO charters, in order to take the
best image quality, regarding the tone and color reproduction, system spatial uniformity, spatial
frequency response, noises, etc. [7].
The digital images obtained by scanning process must be vector or raster. Vector images are made
up of basic geometric shapes such as points, lines and curves. The relationship between shapes is
expressed as a math equation, which allows the image to scale up or down in size without losing its
quality. Raster images are made up of a set grid of dots called pixels where each pixel is assigned a
color value. Unlike a vector image, raster images are resolution dependent.
In the most of the cases for preserving their documental heritage the archival institution use the raster
images over the vector images. For both types of digital images exists a various number of file
formats. The most useful file formats used worldwide for long-term digital preservation as master
copy are TIFF format, JPEG 2000[3], and for access copy JPEG, GIF. The quality of digital raster
images is depending on some technical characteristics such are: Spatial Resolution; Signal Resolution
or Bit- depth; Color Mode, and Compression method, presented below.
4.1. Digital file formats
4.1.1. Tagged Image File Format (TIFF)
The TIFF format was originally developed by the Aldus Corporation, and since 1996 the TIFF
specification was maintained by Adobe Systems Incorporated. The TIFF file is a tag-based file format
for storing and interchanging raster images. It starts with an 8-byte image file header (IFH) that point
to the image file directory (IFD) with the associated bitmap [8]. The IFD contains information about
the image in addition to pointers to the actual image data. The TIFF tags, which are contained in the
header and in the IFDs, contain basic information, the manner in which the image data are organized
and whether a compression scheme is used, for example. The TIFF 6.0 version offers users the option
to use their additional tags. TIFF supports color depths from 1-24 bit and a wide range of compression
types (RLE, LZW, CCITT Group 3 and Group 4, and JPEG), as well as uncompressed data. TIFF
also incorporates the most comprehensive metadata support of any raster format, allowing the
addition of a wide variety of technical and resource discovery information to be included. The TIFF
specifications are freely available for use.
4.1.2. JPEG File Formats
The JPEG standard (ISO/IEC 10918) was created in 1992 (latest version, 1994) as the result of a
process that started in 1986 [9]. It is composed of four separate parts and an amalgam of coding
11
modes. JPEG itself is not a file format, but represent an image compression algorithm. The File
Interchange Format (JFIF) has become a de facto standard; and is commonly referred to as the JPEG
file format. The JPEG algorithm is supported by a number of other raster image formats, including
TIFF. JFIF and SPIFF are 24-bit color formats and use lossy compression. JPEG is particularly suited
to the storage of complex color images, such as photographs.
4.1.3. JPEG 2000 File Formats
JPEG 2000 is an improved version of lossy JPEG algorithm, developed by the ISO JPEG group in
2000. The JPEG 2000 has the option for lossless compression using wavelet compression to achieve
higher compression rates with a lower corresponding reduction in image quality. The JPEG 2000 can
compress a file to a desired size specifying the size in bytes, or specifying a compression ratio. It
supports color depths up to 24-bit (true color) and is best suited to complex color images, such as
photography. JPEG 2000 in the year 2007 is divided into twelve standards that are all more or less
derivations of or supplements to the first standard. A JPEG 2000 file consists of a succession of
boxes. Each box can contain other boxes and have a variable length determined by the first four bytes
and a type that is determined by the second sequence of the four bytes. Every JPEG 2000 file begins
with a JPEG 2000 signature box, followed by a file type box where is determined the type and the
version. This is followed by the header box, which contains the boxes with the resolution, bit depth
and color specifications.
4.2. File format result assessments
The selected test images are scanned with the same resolution in different file format. To decide
whether which resolution can provide the right image quality we referred to equation of Quality Index
(QI). In order to get a quality index QI=8 to a document where the weight of smallest character is 1
mm we tested the resolution on: dpi = 2*8/0.339*0.1 = 400 pixel per inch. Observed by human eyes
it does not distinguish any change to three different types of file format used. In the figure is
illustrated a part of the same document scanned with resolution 300 dpi, RGB 8-bit for channel and
are saved in three different file formats at actual size 1” = 1”.
Figure 1: Examples of fragments (TIFF uncompressed; TIFF compressed and JPEG
2000 lossless)
The LZW compression method is lossless that’s because we have not found any degradation of the
image quality. The image clarity of both JPEG 2000 lossless and 50 quality was high and there was
not any visible defects when evaluate their histogram.
For the same images were calculated their storage capacity. As a result we have found that that TIFF
LZW in lossless mode may save about 49% of storage capacity compared to an uncompressed TIFF.
The images scanned with JPEG 2000 (jpf) in lossless mode has not offer significant benefit compare
with TIFF LZW. The documents saved in JPEG 2000 (jpf) format with 50% quality they occupy
about 80% less storage capacity then JPEG 2000 in lossless mode. In the figure 2 are demonstrated
the results of 5 tested images, scanned with 300dpi resolution and saved in five different file formats.
12
Figure 2: Storage capacity of documents saved in different file formats.
Regarding to the sustainability factors, we employed two different methods for evaluation process
proposed by Rog and van Wijk [11] and the second proposed by Arms and Fleischhauer [12]. Both
methods analyze almost the same sustainability factors but use different ways for scoring. In the first
method the decision way on weighing factor values of different criteria and characteristics was
unclear and influent directly to the assessment values we decided not to use it.
Table 1: A collection of the sustainability factors for both three file formats: TIFF,
JPEG 2000 and JPEG.
Sustainability
Factors
Disclosure
Adaption
Selfdocumentation
Impact of
Patents
Technical
Protection
Mechanisms
Availability of
tools
Dependencies
(hardware&
Software)
Format: TIFF
Format: JPEG 2000(jpf)
Format:
JPEG2
Uncompres
sed
Good
Wide
Adoption.
Negligible
support in
browsers.
Acceptable
Lossless
Lossless
50%
JPEG
Good
Wide
Adoption.
Negligible
support in
browsers.
Acceptable
Good
Moderate
Adoption.
Negligible
support in
browsers.
Good
Good
Moderate
Adoption.
Negligible
support in
browsers.
Good
Good
Wide Adoption
No Impact
Low Impact
No Impact
Little or No
Impact
No Impact
No Impact
No Impact
Little or No
Impact
No Impact
Wide
Availability
Wide
Availability
No
dependencies
Limited to
Moderate
Availability
No
dependencies
Wide
Availability
No
dependencie
s
Limited to
Moderate
Availability
No
dependencies
Acceptable
No Impact
No
dependencies
We adapted the second method but eliminating the quality and functionality factors. The results on
the table highlight that both file formats can be used as digital long term preservation modes but more
acceptable is the TIFF format due to the limited or moderate availability of tools. A number of
important European Institutions, like National Archives of United Kingdom from 2013 have decided
to use the JPEG 2000 format, as their preferred file format for long term preservation.
2
This will be used only as access copy
13
Conclusions
The results reveal the situation that for deciding the file formats that will be used for long term
preservation is necessary to consider same physical aspects of documents (size, original material,
physical conditions) and reason why we need to preserve them. Both file formats TIFF LWZ and
JPEG 2000 Lossless met all technical characteristics of clarity and quality and can be chosen as file
formats for long term preservation.
References
1. Digital Preservation Coalition (2008) "Introduction: Definitions and Concepts". Digital
Preservation Handbook. York, UK. Last retrieved 29 August 2015.
2. Puglia, S., Reed, J., Rhodes, E., U.S. National Archives and Records Administration (2010)
"Technical Guidelines for Digitizing Archival Records for Electronic Access: Creation of Produciton
Master Files – Raster Images”.
http://www.archives.gov/preservation/technical/guidelines.pdf. Last retrieved 05.09.2015.
3. Topi, A., Xhuvani, A. (2015). E-archiving Architecture and Actual Challenges in Albania,
International Conference on Innovation Technologies IN-TECH 2015, September 2015, Croatia,
391-394.
4. Levigo solutions GmbH (2011), “Reproducibility of Archived Documents”
http://www.pdfa.org/organization/levigo-solutions-gmbh; Last retrieved 04 October 2015.
5. https://en.wikipedia.org/wiki/File_format. Last retrieved 2 September 2015.
6. Arms, A., Fleischhauer, C., (2005): “Digital Formats Factors for Sustainability, Functionality and
quality”. IS&T Archiving 2005 Conference, Washington, D.C.
7. Murphy P. E., Rochester Institute of Technology (2002): “A Testing Procedure to Characterize
Color and Spatial Quality of Digital Cameras Used to Image Cultural Heritage”; http://www.artsi.org/PDFs/Metric/EPMurphyThesis05.pdf. Last retrieved 16 September 2015.
8. Part 1 from the TIFF 6.0 specs: http://partners.adobe.com/public/developer/en/tiff/. Last retrieved
22 September 2015.
9. http://www.jpeg.org; Last retrieved 18 September 2015.
10. Kenney, Anne R.; Chapman, Stephen (1995): Digital Resolution Requirements for Replacing
Text-Based Material: Methods for Benchmarking Image Quality.
http://www.clir.org/pubs/reports/pub53/pub53.pdf. Last retrieved 14.09.2015.
11. Franziska, S., Frey, J., & Reilly M. (2006). Digital Imaging for Photographic Collections.
12. Van Wijk, C. & Rog, J. (2007). “Evaluating File Formats for Long-Term Preservation.”
Presentation at International Conference on Digital Preservation, Beijing, China, Oct 11–12.
http://ipres.las.ac.cn/pdf/Caroline-iPRES2007-11-12oct_CW.pdf (accessed September 21, 2015).
14
From Traditional Markets to E-Commerce and Finally to
Social Media Commerce
Ardian Hyseni
Contemporary Sciences and Technologies
South East European University, Skopje, Macedonia
ah16609@seeu.edu.mk
http://www.seeu.edu.mk/
Abstract. These days, getting new customers is much easier than in the past. People and customers
are online sharing and exchanging ideas on products and it has become easier to find products over
the internet and lately; with social media, where people can look for information from reviews and
comments on sites. This way has changed shopping to a social experience and is the key element to
the growth of social commerce. Businesses want to connect with people and customers which they
do business, also they want customers opinions and reviews. By using social media, companies can
now easily create an interaction between the company product and the customer. It is about
connecting them customers directly with your website or even making them: from visitors into loyal
customers. To fulfill research objectives, it is prepared a questionnaire survey through which we test
and analyze the research model and hypothesis. In the conclusion, we discuss about research findings
and the future of social media commerce.
Keywords: Social Media Commerce, Social Media Commerce Application, Social Media
1. Introduction
Social media e-commerce, is a trending way of making business over the internet despite e-commerce
that is made by websites nowadays, social media has made it available to buy and sell products,
through the social networking sites like Facebook twitter etc. Social media isn’t just about posting a
feed on Facebook or twitter, or putting a like button or a comment on your website, but it is about
connecting customers directly with your website and making them visitors, loyal customers. Back in
1994 when Jeff Bezos founded the Amazon.com and in 1995 Pierre Omidyar started the P2P
marketplace on eBay [1]. Both these sites were social networks but these marketplaces were for
products with discount prices. Customers could leave feedback, post reviews and comment for the
products they bought. This was the new era of commerce through the internet. After eBay and
Amazon, in 2004, Facebook was founded by a group of Harvard students. Facebook now is a leading
social networking site for a numbers of users, fans and business pages [2].
Businesses want to connect with people and customers which they do business with, and they want
customers opinions and reviews [7]. By using social media, companies now can easily create an
interaction between company products and the customer. To understand and hear the voice of your
customers, businesses need to keep up with the up to date technology. Social media marketing is
constantly evolving in sites like: Facebook, Twitter and LinkedIn that are leaders of the online
networking, which are the current communication trends [5]. Businesses need to combine new
technologies with the traditional marketing, so to increase the sales revenue [3]. Social media is not
just another way of marketing, but it must become as part of a company, that should be integrated. It
is understandable that businesses should take more seriously the involvement and planning of social
media for commercial gain.
15
Most of the e-commerce sites have started implementing social media networking site services, in
order to improve interaction and increase the active participation of user in order to increase the
revenues [4]. In social networking sites, users can get answers and interact with each other, in order
to get more information for a specific product or even a service.
When a user wants to order a product online, then may ask and find out more information on the
social networking sites. The main aim is to examine and find out at what degree people use social
media commerce, and about the impact of social media in peoples' decision making for purchase or
in ecommerce.
A social networking site consists of large number of users, which are of the potential content of
generators and massive source of information [6]. Users generate new ideas, advertise and add a
value at a little cost, while increasing the effectiveness by understanding customer needs, identifying
potential customers, and building customer loyalty [11].
The increased number of users in social networks has led to a new shopping trend, where people
leverage social networks to make the right purchase. While businesses spend thousands of the money
for the marketing which is temporary on a TV station or in the Newspaper, while in social networking
sites, people who engage to your page may become lifelong loyal customers [10]. Businesses do not
need to pay for advertising in social shopping, they only post products in their business page and all
customers are able to see it[8].
2. Research Methodology
In the research methodology research questions, hypothesis and survey questions will be presented
in order to explore the way which will be taken in order to test and examine the hypothesis and results
of the research, and how the data collected and results will be analyzed.
2.1 Research Methodology
To test the developed model and hypothesis, it is required an effective research and methodology. It
should be considered the conditions, such as research questions, purpose of questions, and the amount
of time [9]. Quantitative and qualitative researches are to explain and understand the phenomena.
The research can be achieved by analyzing experiences and thoughts of individuals and of groups.
Quantitative research questions are the best strategy for gathering data in this surveying case. For
this purpose, survey is the best strategy suggested for gathering data and finding the answers.
Through survey can get a bigger number of participants for a short period of time.
2.2 Research Questions
According to the hypothesis, research questions will be prepared, which will be raised and based on
experiences and expectations of participants in this research. In this research, there will be 25
questions; which all of these have its own importance and each question has its own purpose of
research. Through these prepared questions, will examine and find answers to the research findings.
Results of the answers provided in the survey from participants, will lead us in which social
networking site to be focused, for preparing of the application. Questions will be classified through
a process from which will lead to the adequate answers and steps to be taken to proceed with the
questionnaire. Survey contains three parts of questions. First part contains three questions about
participant’s educational level, department, and the gender they belong to. Second part contains
questions about social media commerce, what knowledge do participants have and what do they
think about future of social media commerce. Third part contains questions about measuring; on how
much impact does social media have in participant’s decision making, and last question about
measuring is; how much do participants believe in that social media commerce, will be a leading way
of commerce in the future.
16
2.3 Hypothesis
In this research, three hypotheses are raised that are going to be examined through some questions in
the survey that will be taken. Results of this Hypothesis will be after analyzing the data collected
from participants in the survey and the questions that are raised will give answers to the hypothesis
raised. The hypotheses raised in this research are:
H1. Social media is an integrative part of people’s everyday life
H2. Social media e-commerce is a new concept that people are not aware of
H3. Selling and buying through social media is a competitive advantage for companies
H3a. Users spend more time on social media, thus they are influenced by social
marketing media
H3b. Active users on social media are more likely to buy if recommended within
the platform of their choice
H4b. The Facebook’s platform offers advantages of having the customers, thus
selling through Facebook is a win-win option for businesses.
By applying the quantitative method of survey questions, we will try to verify these hypotheses in
the upcoming chapter.
2.4 Survey
After preparing the questions, a survey was posted on Facebook and sent to students with the
academic background. Social media remains the best way of sharing information, especially for a
research that is being taken. For a few seconds, you can target friends and colleagues or post the
survey in the Facebook group of your department or university and will get the answers for a few
days to complete the research. The aim of this questionnaire is to collect data, which will be used to
describe and see how much people do know about the social media commerce, and how they are
informed about this new trending way of commerce; which is in ongoing development. Questions
are developed strategically and the questions and its purpose will be shown on the table:
3. Research findings
In this section are presented the research findings from survey taken by 100 participants. Participants
are from the Republic of Kosovo and Macedonia, mostly of them are students from South East
European University, and others are from different Universities around Europe; who are friends and
colleagues from different departments. It will be seen from what department are participants and
what is the gender participation, education level, and their shopping experience and expectations for
social media commerce in the future.
3.1 Results
In the research that was taken, 100 students have participated from different educational levels.
According to results, it shows that students for Bachelor studies were the bigger number of totally
73, followed by 26 students for Master's degree and only 1 PhD. In Table 1 we will see the
participants from which department and what gender they belong to.
17
18
19
3.2 Hypothesis results
Comparing findings and results of the hypotheses test, 2 of our hypotheses were supported and 1 was
not. First hypotheses were:
H1. Social media is an integrative part of people’s everyday life. It is and people are not only the
users, but have become addicted to social media. In figure 1.7 it is shown that social media is not just
a part of entertaining, but also for buying and selling, product discussions and can be used for creating
of professional and business contacts.
H2. Social media commerce is a new concept that people are not aware of, according to Figure 1.8;
surprisingly 62% of our participants claim that have heard the term social media commerce, but only
9% of participants have bought products through social media commerce.
H3. Selling and buying through social media is a competitive advantage for companies, and it is
supported. 70% of the participants believed that social media commerce will be a leading way of
commerce in the future, according to Figure 1.15. This shows that most of the participants believe in
the social media commerce and companies who apply social media commerce, have or will have
advantage over companies who don’t.
H3a. Users spend more time on social media, thus they’re influenced by social media marketing.
This is completely supported by participants according to Figure 1.20, more than 50% agree that
20
if a product is recommended or commented positively by participants, they would be persuaded
to buy a product.
H3b. Active users on social media, are more likely to buy if recommended within the platform of
their choice. According to Figure 1.22 it is supported that the more time people spend in social
media the more they are persuaded to buy, if a product is recommended by a friend most of the
participants agree to buy.
H4b. Facebook’s platform offers advantages of having customers, thus selling through Facebook
is a win-win option for businesses. According to Figure 1.21, it is an advantage if a company uses
Facebook for selling; hash tags makes it possible that one product to be seen by how many people
have chosen hash tag. Social media offers marketing for free and everyone can access it from
everywhere.
Conclusion
Generally social media commerce is an emerging shopping trend that is increasing significantly. So
many apps for selling products through social networking sites are developed for a very short time.
Based on the previous years the rise of social commerce it tends to be the leading way of commerce
in the coming years.
Finally, based on the research and the app that is developed. It can be concluded that research will
have positive impact in people’s daily life in all aspects of entertaining and shopping experiences. It
can be concluded that Facebook users have a lots of fun, while sharing and writing feedbacks about
product and making purchases. Application that is developed for purpose of this thesis tends to be a
good app for small businesses who want to sell products through Facebook.
References
Social Media Explained Untangling the World’s Most Misunderstood Business Trend Mark
W. Schaefer
2. Collier, Marsha (2012-11-08). Social Media Commerce For Dummies (Kindle Locations 1-2).
Wiley. Kindle Edition.
3. Varela, C. (2015-02-03). Facebook Marketing for Business: Learn to create a successful
campaign, Advertise your Business, Strategies to generate traffic and Boost sales (Social
Media)
4. The Effects of Social Media on E-commerce: A Perspective of Social Impact Theory 2012
45th Hawaii International Conference on System Sciences
5. Facebook for Business Owners: Facebook Marketing For Fan Page Owners and Small
Businesses (Social Media Marketing) (Volume 2) Corson-Knowles, Tom
6. Why We Buy: The Science of Shopping--Updated and Revised for the Internet, the Global
Consumer, and Beyond Underhill, Paco
7. The Social Media Bible: Tactics, Tools, and Strategies for Business Success Safko, Lon
8. E-business and e-commerce management: strategy, implementation, and practice / Dave
Chaffey. -- 4th ed.
9. Social Commerce Research: An integrated view. Lina Zhou, Ping Zhang and Hans-Dieter
Zimmermann Last revised: February 15, 2013
10. The Effects of Social Media on E-commerce: A Perspective of Social Impact Theory | 2012
45th Hawaii International Conference on System Sciences
1.
21
11. The Effects of Social Media on E-commerce: A Perspective of Social Impact Theory: KeeYoung Kwahk , Xi Ge
22
New integral transform in Caputo type fractional difference
operator
Artion Kashuri1, Akli Fundo2, Rozana Liko1
Department of Mathematics, Faculty of Technic Sciences, University “Ismail
Qemali", Vlora, Albania
2
Department of Mathematics, Polytechnic University of Tirana, Albania
{artionkashuri1, rozanaliko862}@gmail.com, aklifundo@yahoo.com3
1
Abstract. In this paper, we introduce Caputo type nabla (q,h)-fractional difference operators and
investigate their basic properties and also to show the applicability of this interesting (q,h)- new
integral transform method and its efficiency in solving linear fractional difference equations.
Differential equations with fractional derivative provide a natural framework for the discussion of
various kinds of real problems modeled by the aid of fractional derivative.
Discrete analogues of some topics of continuous fractional calculus have been developed. Finally,
we provide the general solutions in terms of discrete Mittag-Leffler functions.
Keywords: (q,h)- new integral transform, convolution, fractional difference equations, nabla (q,h)fractional integral, nabla (q,h)- fractional derivative.
1. Introduction
(𝑞, ℎ) −new integral transform is derived from the classical Fourier integral and was introduced by
PhD student Artion Kashuri and Associate Professor Akli Fundo to facilitate the process of solving
linear fractional difference equations in the time domain [6], [7]. (𝑞, ℎ) −new integral transform is
defined for functions of exponential order. Fractional calculus deals with the study of fractional order
integrals and derivatives and their applications [1]. Riemann−Liouville and Caputo are kinds of
fractional derivatives which generalize the ordinary integral and differential operators. Differential
equations with fractional derivative provided a natural framework for the discussion of various kinds
of real problems modeled by the aid of fractional derivative. Discrete analogues of some topics of
continuous fractional calculus have been developed. The aim of this paper is to introduce Caputo
type nabla (𝑞, ℎ) −fractional difference operators and investigate their basic properties, and also to
show the applicability of this interesting (𝑞, ℎ) −new integral transform method and its efficiency in
solving linear fractional difference equations. Here, we solve some linear fractional difference
equations involving Caputo type (𝑞, ℎ) −derivatives and provide the general solutions in terms of
discrete Mittag−Leffler functions.
2. Preliminaries
For the convenience of readers, we provide some basic concepts concerning the nabla calculus on
time scales. By a time scale 𝕋 we understand any nonempty, closed subset of reals with the ordering
inherited from reals. Thus the realsℝ, the integersℤ, the natural numbers ℕ, the nonnegative integers
ℕ0 , the h −numbers hℤ = {hk: k ∈ ℤ} with fixed h > 0, and the q −numbers qℕ0 = {qk : k ∈ ℕ0 }
with fixed q > 1 are examples of time scales. For any t ∈ 𝕋, we define the forward and backward
jump operators as σ(t) ≔ inf{s ∈ 𝕋: s > 𝑡} and ρ(t) ≔ sup{s ∈ 𝕋: s < 𝑡}, respectively. The
23
forward and backward graininess functions are defined as μ(t) ≔ σ(t) − t and μ(t) ≔ t − ρ(t),
respectively. By convention, inf{∅} = sup{𝕋} and sup{∅} = inf{𝕋} for these operators. We say that
a point is left−dense if ρ(t) = t and left− scattered if ρ(t) ≠ t. The right− dense and right−scattered
points are defined in a similar manner. A point which is both left− and right−scattered is discrete. If
inf{𝕋} = a0 > −∞, we define 𝕋k ≔ 𝕋\a0 , otherwise 𝕋k ≔ 𝕋. For a function f: 𝕋 → ℝ and a point
t ∈ 𝕋k , we define f ∇ (t) to be a number such that for ϵ > 0 there exists a neighborhood 𝒰 ⊂ 𝕋 of t
which satisfies
|f(ρ(t)) − f(τ) − f ∇ (t)[τ − ρ(t)]| ≤ ϵ|τ − ρ(t)| for all τ ∈ 𝒰 .
If f ∇ (t) is defined for all t ∈ 𝕋k , then the function obtained is called the ∇ −derivative of f. A function
f: 𝕋 → ℝ is called left−dense continuous or ld−continuous provided it is continuous at every
left−dense point in 𝕋, and f(t+) exists for every right−dense point in 𝕋, and the set of ld−continuous
functions is denoted by Cld (𝕋). If f ∈ Cld (𝕋), then there is a function F such that F ∇ (t) = f(t). In this
case, we define the ∇ −integral as
t
∫a f(τ) ∇τ = F(t) − F(a) for all t ∈ 𝕋 .
Clearly, if f: 𝕋 → ℝ is ld−continuous and t ∈ 𝕋k , then
t
∫ f(τ) ∇τ = ν(t)f(t) .
ρ(t)
A function f ∈ Cld (𝕋) is called ν −regressive if 1 + fν ≠ 0 on 𝕋k , and f ∈ Cld (𝕋) is called
positively−regressive if 1 + fν > 0 on 𝕋k . The set of ν −regressive functions and the set of
positively ν −regressive functions are denoted by ℛν (𝕋) and ℛν+ (𝕋), respectively. For simplicity,
we denote by cℛν (𝕋) the set of ν −regressive constants. Let λ ∈ cℛν (𝕋) and s ∈ 𝕋, then the
generalized exponential function êλ (∙, s) on time scale 𝕋 is denoted by the unique solution of the
initial value problem
x ∇ (t) = λx(t), t ∈ 𝕋k
{
x(s) = 1 .
For p ∈ cℛν (𝕋), define circle minus p by
p
.
1 − pν
Then the unique solution of the initial value problem
x ∇ (t) = −λx(t), t ∈ 𝕋k
{
x(s) = 1 .
takes the form ê⊖νλ (∙, s). It is known that the exponential function êf (∙, s) is strictly positive on
[s, ∞[𝕋 provided that f ∈ ℛ+ ([s, ∞[𝕋 ). The definition of the generalized monomials ĥn : 𝕋 × 𝕋 →
ℝ (n ∈ ℕ0 ) is given by
1,
n=0
⊖ν p ≔ −
ĥn (t, s) = {
t
∫ ĥn−1 (τ, s)∇τ,
n∈ℕ.
s
for s, t ∈ 𝕋, [2]. If we let ĥ∇n (t, s) denote for each fixed s ∈ 𝕋 the derivatives of ĥ∇n (t, s) with respect
to t, then
ĥ∇n (t, s) = ĥn−1 (t, s) for n ∈ ℕ, t ∈ 𝕋k .
An important relation between the generalized exponential function and the monomials is given by
k̂
êλ (t, s) = ∑∞
k=0 λ hk (t, s) for s, t ∈ 𝕋 with t ≥ s ,
(𝕋).
where λ ∈ ℛc
Definition 2.1. The nabla new integral transform ([6]-[8]) of a function f: 𝕋 → ℝ is defined by
1 ∞
𝛫∇ {𝑓}(𝑧, 𝑣) ≔ ∫𝑣 ê⊖ 1 (ρ(τ), z)𝑓(𝜏)∇ 𝜏 for 𝑧 ∈ 𝐷 ,
𝑣
ν 𝑣2
where 𝐷 consists of all complex numbers 𝑧 ∈ ℂ for which the improper ∇ −integral exists.
24
Definition 2.2. The convolution of two functions f, g: 𝕋 → ℝ is defined by
𝑡
(𝑓 ∗ 𝑔)(𝑡) ≔ ∫ 𝑓̂ (𝑡, 𝜌(𝑠))g(s)∇s,
t∈𝕋,
𝑎
where 𝑓̂ is the shift of 𝑓 introduced in [3]. Then
𝛫∇ {𝑓 ∗ 𝑔}(𝑧, 𝑣) = 𝑧 ∙ 𝛫∇ {𝑓}(𝑧, 𝑣) ∙ 𝛫∇ {𝑔}(𝑧, 𝑣) .
The following result is the nabla version of the result obtained in [4] (change of integration order).
Let 𝑓 ∈ 𝐶𝑙𝑑 (𝕋2 ), then
𝑡 𝜂
𝑡 𝑡
∫𝑠 ∫𝑠 𝑓(𝜂, 𝜁) ∇ζ∇η = ∫𝑠 ∫𝜌(𝜁) 𝑓(𝜂, 𝜁) ∇η∇ζ for 𝑠, t ∈ 𝕋 .
3 (𝒒, 𝒉) −Fractional Calculus
Consider the following (𝑞, ℎ) −time scale:
ℎ
𝑡0
𝕋(𝑞,ℎ)
= {𝑡0 𝑞𝑘 + [𝑘]𝑞 ℎ: 𝑘 ∈ ℤ}⋃ {
},
1−𝑞
for 𝑡0 > 0, 𝑞 ≥ 1, ℎ ≥ 0 and 𝑞 + ℎ > 1. Note that if 𝑞 = 1, then the cluster point ℎ⁄1 − 𝑞 = −∞ is
not involved in 𝕋. The forward and backward jump operator is the linear function σ(t) = qt + h and
ρ(t) = 𝑞 −1 (𝑡 − ℎ), respectively. Similarly, the forward and backward graininess is given by μ(t) =
(q − 1)t + h and ν(t) = 𝑞−1 𝜇(𝑡), respectively. Observe that
𝜎 𝑘 (𝑡) = 𝑞𝑘 𝑡 + [𝑘]𝑞 ℎ and 𝜌𝑘 (𝑡) = 𝑞−𝑘 (𝑡 − [𝑘]𝑞 ℎ) .
The following relation ν(𝜌𝑘 (𝑡)) = 𝑞−𝑘 ν(t) holds for t ∈ 𝕋. The nabla (𝑞, ℎ) −derivative of the
𝑡0
function f: 𝕋(𝑞,ℎ)
→ ℝ is defined by
𝑓(𝑡) − 𝑓(𝜌(𝑡)) 𝑓(𝑡) − 𝑓(𝑞̃(𝑡 − ℎ))
=
,
(1 − 𝑞̃)𝑡 + 𝑞̃ℎ
ν(t)
𝑡
0
where 𝑞̃ = 𝑞 −1 . Let t, a ∈ 𝕋(𝑞,ℎ)
such that ℎ⁄1 − 𝑞 ≤ 𝑎 ≤ 𝑡 and f: 𝕋 → ℝ. Then the nabla
(𝑞, ℎ) −integral exists and can be calculated (provided 𝑡 > 𝑎) via the formula
∇(𝑞,ℎ) 𝑓(𝑡) ≔
𝑡
−1
𝑎∇(𝑞,ℎ) 𝑓(𝑡)
𝑛−1
≔ ∫ 𝑓(𝜏)∇τ = ((1 −
𝑞 −1 )𝑡
𝑎
+ 𝑞 −1 ℎ)
∑ 𝑞 −𝑘 𝑓(𝑞 −𝑘 𝑡 + [−𝑘]𝑞 ℎ) .
𝑘=0
𝑡
0
The Taylor monomials and the power functions on 𝕋(𝑞,ℎ)
have the forms
ℎ̂𝑛 (𝑡, 𝑠) =
𝑗
∏𝑛−1
𝑗=0 (𝜎 (𝑡)−𝑠)
[𝑛]𝑞 !
=
𝑗
∏𝑛−1
𝑗=0 (𝑡−𝜌 (𝑠))
(𝑛)
𝑗
where (𝑡 − 𝑠)(𝑞̃,ℎ) = ∏𝑛−1
𝑗=0 (𝑡 − 𝜌 (𝑠)) .
[𝑛]𝑞 !
𝑡
0
Respectively, and the extension of the monomials ℎ̂𝑛 (𝑡, 𝑠) corresponding to 𝕋(𝑞,ℎ)
takes the form
(𝛼)
ℎ̂𝛼 (𝑡, 𝑠) =
(𝑡 − 𝑠)(𝑞̃,ℎ)
Γ𝑞̃ (𝛼 + 1)
,
𝛼∈ℝ.
𝑡
0
Lemma 3.1. [5] Let 𝑚 ∈ ℤ+ , 𝛼 ∈ ℝ, s, t ∈ 𝕋(𝑞,ℎ)
and 𝑛 ∈ ℤ+ , 𝑛 ≥ 𝑚 be such that 𝑡 = 𝜎 𝑛 (𝑠).
Then
ℎ̂𝛼−𝑚 (𝑡, 𝑠),
𝛼 ∉ {0,1, … , 𝑚 − 1},
̂
∇𝑚
(3.1)
(𝑞,ℎ) ℎ𝛼 (𝑡, 𝑠) = {
0,
𝛼 ∈ {0,1, … , 𝑚 − 1} .
0
Let 𝛼 ∈ 𝕋(𝑞,ℎ)
, 𝑎 > ℎ⁄(1 − 𝑞) be fixed. We consider the following restricted (𝑞, ℎ) −time scale:
𝑡
𝑖
̃ 𝜎 (𝑎) = {𝑡 ∈ 𝕋, t ≥ 𝜎 𝑖 (𝑎)},
𝕋
(𝑞,ℎ)
𝑖 = 0,1,2, … ,
(3.2)
25
where the symbol 𝜎 𝑖 stands for the 𝑖th iterate of 𝜎 (analogously, we use the symbol 𝜌𝑖 ). Now we can
continue with the introduction of (𝑞, ℎ) −fractional integral and (𝑞, ℎ) −fractional derivative of a
𝑖
𝑖 (𝑎)
̃ 𝜎 (𝑎) → ℝ. Let 𝑡 ∈ 𝕋
̃ 𝜎(𝑞,ℎ)
function f: 𝕋
.
(𝑞,ℎ)
Definition 3.2. [5] The nabla (𝑞, ℎ) −fractional integral (in the sense of Riemann-Liouville) of order
̃ 𝑎(𝑞,ℎ) is defined by
𝛼 ∈ ℝ+ over the time scale interval [𝑎, 𝑡]⋂𝕋
t
−𝛼
𝑎∇(𝑞,ℎ) 𝑓(𝑡)
= ∫ ĥα−1 (t, ρ(τ))f(τ)∇τ .
(3.3)
a
The nabla (𝑞, ℎ) −fractional derivative (in the sense of Caputo) of order 𝛼 ∈ ℝ+ is defined by
t
𝐶 ∇𝛼
𝑎 (𝑞,ℎ) 𝑓(𝑡)
=
𝑎
∇−(𝑚−𝛼) ∇𝑚 𝑓(𝑡)
= ∫ ĥm−α−1 (t, ρ(τ))∇𝑚 f(τ)∇τ ,
(3.4)
a
where 𝑚 ∈ ℤ+ is such that 𝑚 − 1 < 𝛼 ≤ 𝑚.
4. Application
𝑡
0
In this section we give solutions for fractional initial value problems on 𝑎 ∈ 𝕋(𝑞,ℎ)
.
𝑠,𝜆
Definition 4.1. Let 𝛼, 𝛽, 𝛾 ∈ ℝ. The (𝑞, ℎ) −Mittag-Leffler function 𝐸𝛼,𝛽
(𝑡) is defined by
∞
𝑠,𝜆
(𝑡)
𝐸𝛼,𝛽
=
∞
∑ 𝜆𝑘 ℎ̂𝛼𝑘+𝛽−1 (𝑡, 𝑠)
𝑘=0
(=
∑ 𝜆𝑘
𝑘=0
(𝛼𝑘+𝛽−1)
(𝑡 − 𝑠)(𝑞̃,ℎ)
Γ𝑞̃ (𝛼𝑘 + 𝛽)
),
(4.1)
𝜎 (𝑎)
̃ (𝑞,ℎ) and 𝑡 ≥ 𝑠. It is easy to check that the series on the right-hand side converges
for 𝑠, 𝑡 ∈ 𝕋
𝛼
(absolutely) if |𝜆|(ν(t)) < 1. Now we give (𝑞, ℎ) −new integral transforms of fractional nabla
𝑡0
𝑡0
integral and Caputo nabla derivative on 𝕋(𝑞,ℎ)
. For 𝑡, 𝑠 ∈ 𝕋(𝑞,ℎ)
and 𝛼 ∈ ℝ+ , we have
̂(𝑞,ℎ) {ℎ̂𝛼 (𝑡, 𝑣)}(𝑧) = 𝑧 2𝛼+1 .
𝐾
𝑠,𝜆
(𝑡), we have
From the definition of (𝑞, ℎ) −Mittag-Leffler function 𝐸𝛼,𝛽
∞
𝑎,𝜆
̂(𝑞,ℎ) {𝐸𝛼,𝛽
̂(𝑞,ℎ) {∑ 𝜆𝑘 ℎ̂𝛼𝑘+𝛽−1 (∙, 𝑎)} (𝑧)
(𝑡)}(𝑧) = 𝐾
𝐾
𝑘=0
𝑘 ̂
̂
= ∑∞
𝑘=0 𝜆 𝐾(𝑞,ℎ) {ℎ𝛼𝑘+𝛽−1 (∙, 𝑎)}(𝑧) =
𝑧2𝛽−1
provided |𝑧 2𝛼 𝜆| < 1 .
1−𝜆𝑧2𝛼
(4.2)
𝜎 𝑖 (𝑎)
For 𝛼 = 1, 𝛽 = 1 and 𝑡 ∈ 𝕋(𝑞,ℎ) , we have
𝑎,𝜆
(𝑡) = 𝑒̂𝜆 (𝑡, 𝑎) .
𝐸1,1
Hence
𝑎,𝜆
̂(𝑞,ℎ) {𝑒̂𝜆 (𝑡, 𝑎)}(𝑧) = 𝐾
̂(𝑞,ℎ) {𝐸1,1
(𝑡)}(𝑧) =
𝐾
(4.3)
𝑧
.
1 − 𝜆𝑧 2
Theorem 4.2. The (𝑞, ℎ) −new integral transform of fractional nabla integral is given by
2𝛼 ∙ 𝐾
̂(𝑞,ℎ) { 𝑎∇−𝛼
̂(𝑞,ℎ) {𝑓}(𝑧) .
𝐾
(𝑞,ℎ) 𝑓}(𝑧) = 𝑧
(4.4)
(4.5)
Proof. By convolution, one may write
26
t
−𝛼
̂
̂
𝑎 ∇(𝑞,ℎ) 𝑓(𝑡) = ∫ hα−1 (t, ρ(τ))𝑓(τ)∇τ = (hα−1 (∙, 𝑎) ∗ 𝑓)(𝑡) .
(4.6)
a
Thus
2𝛼 ∙ 𝐾
̂
̂(𝑞,ℎ) { 𝑎∇−𝛼
̂
̂
̂(𝑞,ℎ) {𝑓}(𝑧) .
𝐾
(𝑞,ℎ) 𝑓}(𝑧) = 𝑧 ∙ 𝐾(𝑞,ℎ) {hα−1 (𝑡, 𝑎)}(𝑧) ∙ 𝐾(𝑞,ℎ) {f}(𝑧) = 𝑧
Theorem 4.3. The (𝑞, ℎ) −new integral transform of Caputo nabla derivative is given by
𝑚−1 𝑘
̂
{𝑓}(𝑧)
∇(𝑞,ℎ) 𝑓(𝑎)
𝐾
̂(𝑞,ℎ) { 𝐶𝑎∇𝛼(𝑞,ℎ) f}(𝑧) = (𝑞,ℎ)
𝐾
−
∑ 2(𝛼−𝑘)−1 ,
2𝛼
𝑧
𝑧
(4.7)
𝑘=0
where 𝑚 − 1 < 𝛼 ≤ 𝑚. In particular, if 0 < 𝛼 ≤ 1, then
̂(𝑞,ℎ) { 𝐶𝑎∇𝛼(𝑞,ℎ) f}(𝑧) =
𝐾
̂(𝑞,ℎ) {𝑓}(𝑧) 𝑓(𝑎)
𝐾
− 2𝛼−1 ,
𝑧 2𝛼
𝑧
(4.8)
where ∇0(𝑞,ℎ) 𝑓(𝑎) = 𝑓(𝑎), 𝑘 = 0 and 𝑚 = 1. As an application, we apply (𝑞, ℎ) −new integral
transform method to derive explicit solutions to the homogeneous equations of the form
𝐶 ∇𝛼
𝑑𝑘 = ∇𝑘 𝑦(𝑎), 𝑘 = 0, … , 𝑚 − 1 ,
(4.9)
𝑎 (𝑞,ℎ) y(t) − λy(t) = 0,
𝑚
̃ 𝜎(𝑞,ℎ)(𝑎) (𝑚 ∈ ℕ), 𝑚 − 1 < 𝛼 ≤ 𝑚, λ ∈ ℝ, in terms of the (𝑞, ℎ) −Mittag-Leffler
where 𝑡 ∈ 𝕋
functions. The following statement holds.
Theorem 4.4. Let 𝑚 ∈ ℕ be given by 𝑚 − 1 < 𝛼 ≤ 𝑚 and λ ∈ ℝ. Then the functions
𝑎,𝜆
(𝑡), 𝑖 = 0,1, … , 𝑚 − 1 ,
𝑦𝑖 (𝑡) = 𝐸𝛼,𝑖+1
(4.10)
yield the fundamental system of solutions to equation (4.9).
Proof. Applying (𝑞, ℎ) −new integral transform to (4.9) and taking (4.7) into account, we have
𝑚−1
̂(𝑞,ℎ) {𝑦}(𝑧) = ∑ 𝑑𝑖
𝐾
𝑖=0
𝑧 2𝑖+1
.
1 − 𝜆𝑧 2𝛼
(4.11)
𝑧 2𝑖+1
.
1 − 𝜆𝑧 2𝛼
(4.12)
Formula (4.2) with 𝛽 = 𝑖 + 1 yields
𝑎,𝜆
̂(𝑞,ℎ) {𝐸𝛼,𝑖+1
(𝑡)}(𝑧) =
𝐾
Thus, relation (4.11) yields
𝑎,𝜆
𝑦(𝑡) = ∑𝑚−1
𝑖=0 𝑑𝑖 𝑦𝑖 (𝑡), where 𝑦𝑖 (𝑡) = 𝐸𝛼,𝑖+1 (𝑡) .
∎
Acknowledgments. Express our sincere gratitude to Associate Professor Akli Fundo,
whose suggestion led to conduction of this research.
References
27
1. M.R.S. Rahmat, M.S.M. Noorani: Caputo type fractional difference operator and its application
on discrete time scales, Advanced in Difference Equations, 2015:160 DOI 10.1186/s13662-0150496-5, (2015).
2. B.M. Peterson, A (eds.): Advanced in Dynamic Equations on Time Scales, Birkh𝑎̈ user, Boston,
(2003).
3. Kisela, T: Power functions and essentials of fractional calculus on isolated time scales, Adv. Differ.
Equ., Article ID 259, (2013).
4. Karpuz: Volterra theory on time scales, Results Math., 65, 263-292, (2014).
5. Cermak, J, Kisela, T, Nechvatal, L: Discrete Mittag-Leffler functions in linear fractional
Difference equations, Abstr. Appl. Anal., Article ID 565067, (2011).
6. A. Kashuri, A. Fundo: A New Integral Transform, Advanced in Theoretical and Applied
Mathematics, ISSN 0973-4554, Vol. 8, No. 1, 27-43, (2013).
7. A. Kashuri, A. Fundo, M. Kreku: Mixture of a New Integral Transform and Homotopy
Perturbation Method for Solving Nonlinear Partial Differential Equations, Advances in Pure
Mathematics, Vol. 3, No. 3, 317-323, (2013).
8. A. Kashuri, A. Fundo, R. Liko: New Integral Transform for Solving Some Fractional Differential
Equations, International Journal of Pure and Applied Mathematics, ISSN 1311-8080, Vol. 103, No.
4, 675-682, (2015).
28
The mobile telecom operator T-Mobile Macedonia AD
Skopje abuses its dominant position harmful for the mobile
telecom operators ONE and VIP in the telecom market
Boris Arsov
Agency for electronic communication
boris.arsov@aec.mk
Abstract. T-Mobile’s abusing of its dominant position is referring to the postpaid tariff models Relax
and Relax Surf regarding the residential customers and the postpaid tariff models Business and
Business Surf regarding its business customers. The prominent price offer to consumers is a case of
predatory prices squeeze identified as a way of preventing or restricting competition in theory and
practice. T-Mobile provides services at unreasonably low prices, prices below the level of
expenditures necessary for their provision, as it is specific case with the postpaid tariff models Relax
and Relax Surf regarding the residential customers and the postpaid tariff plans Business and
Business Surf regarding the business customers. Providing predatory prices is anti-competitive
commercial strategy used by a certain dominant enterprise by dumping its prices on a certain relevant
market to a price level forcing its competitors to leave the market. The interconnection costs or the
cost amount of call termination in their own network or in other operator’s network regardless if it is
fixed or mobile are determined by the Agency for Electronic Communications. The major element
of the market economy is the free and effective competition. There is not a market economy without
a competition as there is not a competition without a market economy. The competition in the market
is a simple and effective means ensuring that the products and the services are offered to the
customers with an excellent quality and competitive prices.
By providing such services with predatory prices, T-Mobile intends to discipline its competitors, to
protect and enhance its extended dominant market power on a long-term period. Disabling the
competitiveness of the other operators and the inability to replicate the offer provided by an operator
with significant market power is undoubtedly a risk leading to competition’s elimination on a long
term. Thus, T-Mobile destroys the competition in the field of mobile telephony, taking advantage of
its significant market power and dominant market position of the two entities, by providing
conditions to which the other market participants are not able to respond due to the expenditure
services structure. The competition and the free markets are the main engines of productivity,
efficiency, product development, innovation and appropriate pricing. The competitive markets
stimulate better technologies and technological development in order to provide products and
services to their customers with high quality and prices reflecting the efficient producers’
expenditures.
Keywords: mobile operator, Macedonia, agency for electronic communication, tariff models
1. Introduction
The prominent price offer to consumers is a case of predatory prices squeeze identified as a way of
preventing or restricting competition in theory and practice. T-Mobile provides services at
unreasonably low prices, prices below the level of expenditures necessary for their provision, as it is
specific case with the postpaid tariff models Relax and Relax Surf regarding the residential customers
and the postpaid tariff plans Business and Business Surf regarding the business customers.
29
Providing predatory prices is anti-competitive commercial strategy used by a certain dominant
enterprise by dumping its prices on a certain relevant market to a price level forcing its competitors
to leave the market. The interconnection costs or the cost amount of call termination in their own
network or in other operator’s network regardless if it is fixed or mobile are determined by the
Agency for Electronic Communications. The major element of the market economy is the free and
effective competition. There is not a market economy without a competition as there is not a
competition without a market economy. By providing such services with predatory prices, T-Mobile
intends to discipline its competitors, to protect and enhance its extended dominant market power on
a long-term period. Disabling the competitiveness of the other operators and the inability to replicate
the offer provided by an operator with significant market power is undoubtedly a risk leading to
competition’s elimination on a long term. Thus, T-Mobile destroys the competition in the field of
mobile telephony, taking advantage of its significant market power and dominant market position of
the two entities, by providing conditions to which the other market participants are not able to respond
due to the expenditure services structure. The competition and the free markets are the main engines
of productivity, efficiency, product development, innovation and appropriate pricing. The
competitive markets stimulate better technologies and technological development in order to provide
products and services to their customers with high quality and prices reflecting the efficient
producers’ expenditures.
2. Facts Statement and Properly Made Calculations and Conclusions
The Commission for Protection of Competition is an institution responsible for implementing the
Law on Protection of Competition. The competition right has been established for the protection of
free competition between the enterprises in the market. The free competition is crucial for all
countries whose economies are based on the principle of a free market, where the funds allocation is
a result of the supply and demand relationship in the market, but not a result of the measures the
country intervenes within the relations between the enterprises. The application of the competition
rules is aimed at establishing a market where the enterprises are equal under equal conditions and
their position in the market to be valued according to the quality of the goods and services they offer.
T-Mobile Macedonia is the first mobile operator and market leader in mobile telephony in
Macedonia. The company was founded in 1996 and has been part of the Deutsche Telekom
Group since 2006. T-Mobile is the largest mobile operator in Macedonia. Relax, Relax Surf,
Business and Business Surf are postpaid tariff models offered on the market of mobile
telecommunication services by T-Mobile Macedonia to their customers and ultimate users by
concluding a contract for using them; for new customers the contract is for a period of 24 months
while for already existing customers the contractual terms are extended for 24 months.
3. Relax Surf postpaid tariff model offers to its residential customers:

Unlimited calls within T-Mobile and T-Home customers;

Unlimited 3G internet;

Free minutes for calls within the other networks;

Free SMS with Relax Surf L;

Wireless Wi-Fi internet on 500 Telecom HotSpot locations.
The newly Relax Surf tariff models are available for all existing and new customers. 3
Table 1 shows the prices and specifications of the offered services which will be subject to the
analysis later in the finding provided by the official website of the company T-Mobile Macedonia.
(1 EUR = 61.5 MKD; 1 MKD = 0.016 EUR = 1.6 cent)
3
Website: http://t-mobile.mk/public/relax-surf-new.nspx, seen on 20.10.2013.
30
Table 1: Prices of the offered mobile services provided from the website of T-Mobile: http://tmobile.mk/public/relax-surf-new.nspx
Relax
Relax
Relax
Relax
Surf S
Surf M
Surf L
Surf XL
Monthly fee (MKD)
Included minutes in T-Mobile and
T-home
Included internet
Included minutes in other networks
Price per minute in all network
after exceeding the included
minutes
399 den
100 min
599 den
1000 min
1199 den
2000 min
1799 den
3000 min
300 MB
/
5, 9
den/min
1GB
/
4,9
den/min
2GB
/
3,9
den/min
3GB
400 min
2,9
den/min
Calculations and conclusions
The offered prices shown above represent a case of predatory price squeeze identified as a way of
preventing or restricting competition in theory, which according to Article 11, paragraph 2 item 1 of
the Law on Protection of Competition is an abuse of a dominant market position by the company TMobile Macedonia.
Offering predatory prices is an anti-competitive commercial strategy with which a specific enterprise
or company on a specific relevant market lowers prices to that price level thus disturbing the normal
development of the relevant market, wherein the other participants or its competitors are forced to
leave that relevant market because of unequal competition and by failing to offer services at realistic
prices.
If in the determination of the relevant market is going a step further in the direction of detailing or
clarifying the data for the relevant mobile market based on the customers’ tariff models, then the
mobile telephony service market can be considered as composed of post-paid and pre-paid element.
The facts and data relating to the share the company T-Mobile Macedonia AD owns on the relevant
market for mobile telephony services in terms of both the number of active customers and the
revenues from mobile services undoubtedly refer to meeting the requirement for a dominant position
of T-Mobile Macedonia on the relevant market for mobile telephony services.
T-Mobile Macedonia offers its products and / or services at unreasonably low prices
The cost analysis (calculations) shows that these tariffs are offered at unreasonably low prices. The
volume of traffic included within the tariff models monthly fee vastly contributes to the cost structure
of the tariff model and can often be the deciding factor in the offer creation which is mainly a
predatory price-oriented. The interconnection costs or the amount of the cost of call termination in
their own or another network, fixed or mobile, are determined by the regulatory body for electronic
communications, the Agency for Electronic Communications.
According to article 48 of the Law on Electronic Communications (Official Gazette, 13/2005,
14/2007, 55/2007, 98/2008, 83/10, and 13/12) and according to the results of the third analysis
Market 12 - call termination in public mobile communications networks, or the report on the
calculation of prices for call termination in public mobile communications networks in the country
based on LRIC methodology, the Agency for Electronic Communications stipulates that the
interconnection cost for call termination in the network of T-Mobile Macedonia coming from another
network or their own network, is 3 denars per minute, while the interconnection cost for call
termination on the network of VIP Operator ONE LTD and ONE telecommunication services LTD
31
Skopje is 4 denars per minute.4 Additionally, according to the Reference Interconnection Offer of
Macedonian Telekom (MATERIO) the compensation for service on regional call termination per
minute in period with normal tariff is 0.45 MKD/minute or 0.2 MKD/minute for regional service call
termination per minute in period with cheap tariff. 5 The Agency for Electronic Communications has
introduced the symmetry principle regarding the determination of the interconnection compensation
amount in the fixed networks. This means that the compensation for a service of regional call
termination per minute in period with normal tariff from 0.45 MKD/minute or 0.20 MKD/minute
applies to calls termination in any fixed operator network. In order to provide a better illustration of
the model for costs calculation, it will be applied and elaborated in detail on the postpaid tariff model
Relax Surf XL. (Table 1, last column)
The basic characteristics of the tariff model Relax Surf XL are:

The monthly fee is 1.799 denars with included VAT;

The price includes 3.000 free minutes for calls within T-Mobile Macedonia and T-Home
or Macedonian Telekom networks.

The price includes 400 minutes for call within other networks, in other words fixed
networks or other mobile operator networks.
(1 EUR = 61,5 MKD; 1 MKD = 0,016 EUR = 1,6 cent).
In order to provide a proper review of the actual market situation and the behavior of the Macedonian
user of mobile communication services, the common cases in practice will be explained in the costs
calculation of the cost.
1) Case num.1:
The overall traffic from 3000 free minutes for calls within the networks of T-Mobile Macedonia and
Macedonian Telekom is realized in the network of Macedonian Telecom, while the overall traffic of
400 free minutes to other networks is realized in the networks of mobile operators ONE
telecommunication Services LTD Skopje and VIP Operator LTD Skopje.
The costs calculated for providing the service ae:

The cost for providing the service, included minutes within T-Mobile Macedonia and
Macedonian Telekom, where all traffic is realized in the network of Macedonian
Telecom is 1328.00 denars. If it is taken an average that 70% of the calls are calls in
normal tariff while 30% are in cheap tariff, the cost for providing the service 3,000 free
minutes for calls in the network of Macedonian Telecom is calculated as:
((0,3*0,2)*3.000 min +(0,7*0,45)*3.000 min )*1,18 VAT);

The cost for providing the service, included minutes to other networks where the overall
traffic of 400 free minutes is realized in the network of mobile operators ONE
Telecomunication services LTD Skopje and VIP Operator LTD Skopje is 1888.00 denars
and it is calculated by the formula: (400 min * 4 den *1.18 VAT);

The total cost of T-Mobile Macedonia for providing the tariff is amounted to 1.328 +
1.888 = 3216.00 denars. On the other hand, the amount of the monthly fee for the tariff
model Relax XL is 1799.00 denars, so it is certainly concluded that the costs for
providing the tariff Relax XL are only 1417.00 denars higher than the monthly fee.
Table 2: Example for a Calculation of the package cost Relax 6
Cost for the service providing
Included minutes within Т-Mobile and Т-Home (The overall
traffic is realized in the T-Home network)
4
MKD
1,328
Source: website of the Agency for Electronic Communications: www.aec.mk ;
5 Source: Reference Interconnection Offer of Macedonian Telekom AD Skopje (MATERIO),
website
of Macedonian Telekom AD Skopje: www.telekom.mk
6 Source: Initial data are taken from the website and the offer of T-Mobile Ad Skopje: www.tmobile.mk
32
Included minutes within other networks (Interconnection) The
overall volume of traffic is realized in ONE and VIP operator
networks
Included internet
Total
Margin
1,888
0
3,216
- 1,417
2) Case num.2
The overall traffic from 3000 free minutes for calls within the networks of T-Mobile Macedonia and
Macedonian Telekom is realized in the network of Macedonian Telecom, while the overall traffic of
400 free minutes to other networks is realized in the networks of other fixed operators.
The costs calculated for providing the service are:

The cost for providing the service, included minutes to T-Mobile Macedonia and
Macedonian Telecom, where all traffic is realized in the network of T-Mobile Macedonia
is 10,620.00 denars or (3.000 min * 3 den * 1.18 VAT);

The cost for providing the service, included minutes to other networks where the overall
traffic of 400 free minutes is realized in networks of other fixed operators is 177,00
denars and it is calculated by the formula
((0,3*0,2)*400 min+(0,7*0,45)*400 min
)*1,18 VAT);
 The total cost of T-Mobile Macedonia for providing this tariff is amounted to 10 620 +
177 = 10,797.00 denars. On the other hand, the amount of the monthly fee for the tariff
model Relax Surf XL is 1,799.00 denars, so it is certainly concluded that the costs for
providing the tariff Relax XL is only 8,998.00 denars higher than the monthly fee. (Table
3).
It must be noted that the calculation model does not incorporate and does not take into consideration
the cost for providing the service included Internet, which further increases the overall level of cost
which becomes even more evident in the offer of services at unreasonably low and far anticompetitive prices by T-Mobile Macedonia. Given the results of the cost calculation of the model it
is more than evident that the cost for providing these analyzed tariffs are significantly higher than for
their monthly fee, so the offer of services at unreasonably low prices is more evident and clear.
Table 3: Example for a Calculation of the package costs Relax 7
Cost for the service providing
Included minutes within Т-Mobile and Т-Home (The overall traffic
is realized in the T-Home network)
Included minutes within other networks (Interconnection) The
overall volume of traffic is realized in other fixed operator networks
Included internet
Total
Margin
MKD
10,620
177
0
10,797
- 8,998
Conclusion
Considering the above facts about the size of the market share of T-Mobile Macedonia and the offered
unreasonably low prices for mobile telephony service, it is more than clear that there is an abuse of
dominant position by offering predatory prices. T-Mobile Macedonia offers low prices intending to
7
Source: Initial data are taken from the website and the offer of T-Mobile: www.t-mobile.mk
33
eliminate and discipline its competitors in the market or to prevent their entry in order to protect and
increase its market power on a long-term period. The major element of the market economy is the
free and effective competition. There is not a market economy without a competition as there is not
a competition without a market economy. The competition in the market is a simple and effective
means ensuring that the products and the services are offered to the customers with an excellent
quality and competitive prices. By providing such services with predatory prices, T-Mobile intends
to discipline its competitors, to protect and enhance its extended dominant market power on a longterm period. Disabling the competitiveness of the other operators and the inability to replicate the
offer provided by an operator with significant market power is undoubtedly a risk leading to
competition’s elimination on a long term. Failing to take the appropriate measures in accordance with
the Law on Protection of Competition, or their prolonging means enabling the dominant behavior of
the mobile service market by T-Mobile Macedonia can cause serious distortion of competition in the
market and strengthen the dominant position of T -Mobile Macedonia.
34
References:
1.
2.
3.
4.
5.
http://t-mobile.mk/public/relax-tarifa-new.nspx;
http://t-mobile.mk/public/relax-surf-new.nspx;
Agency for electronic communications, Final Document of the third market analysis 12 - the
market for call termination in public mobile communications networks; 2013;
Agency for electronic communications, Reference Interconnection Offer of Macedonian
Telekom AD Skopje (MATERIO); 2013;
Law for Electronic Communications (Official Gazette, 13/2005, 14/2007, 55/2007, 98/2008,
83/10, and 13/12); 2012.
35
The Analytic Hierarchic Process: A Model to Evaluate the
Risk of Getting Cancer
Daniela Qendraj (Halidini)1, Markela Muca2
1
Department of Mathematics, Faculty of Information Technology, University
“Aleksandër Moisiu” Durrës, Albania.
2
Department of Mathematics, Faculty of Natural Sciences, UT, Albania.
daniela_qendraj@hotmail.com1, markela.moutsa@yahoo.com2
Abstract. The aim of this study is to classify the risk factors of being with a type of cancer by
applying the AHP (analytic hierarchy process) scale. The Saaty’s analytic hierarchy process is
applied for computing weights for the factors, is a modeling for classification. A case study of QSUT
Tirana hospital data base is presented to rank four cancer types. By the data collect from the hospital
during the years 2011-2015, the results show that lung cancer is mostly spread in humans, and the
prostate cancer is lower spread in humans.
Keywords: AHP process, local weights, IC, cancer types.
1. Introduction
Now days the most discussions are focused in developing of science in cancer disease. Despite of the
long searches, the factors of causing cancer are several and not fixed. Science has concluded that the
number of factors that causes cancer depends on the type of cancer. For example nasal carcinoma is
caused by smoking, working under poor ventilation, herbal drugs, nasal balms [6]. The most factors
leading to cancer are the chemical exposure, family history, alcohol. So, there are no clear evidences
to provide specific risk factor to cancer. To classify the factors in cancer disease we have the model
of Multiple Criteria Decision Making that is analytic hierarchy process (AHP). The AHP provides a
structured problem, to evaluate alternative solutions, and is presented firstly by Thomas Saaty in
1971-1975 [1]. It is used to derive ratio scales from both discrete and continuous paired comparisons.
In the discrete case these comparisons leads to dominance matrices and in the continuous case to
kernels of Fredholm operators, for which ratio scales are derived in the form of principal
eigenvectors.
2. Analytic Hierarchy Process (AHP)
The analytic hierarchy process is a tool in multi-criteria decision making which compose the factors
in a hierarchic problem structure. The applications are in different research areas especially in the
government, health [6], business [7], industry, building and education [4 ].
At first of the AHP procedures, is the aim of the problem, represented by the decision maker. It
considers the first level of the hierarchy, second level are the multiple criteria, the last level are
alternatives. The general steps of the AHP model are as follow:
1. Construct a hierarchy structure for the problem.
36
Table 1 Saaty fundamental scale
Preference
for
pair-wise Preference
comparison
number
Equally important
1
Moderately more important
3
Strongly more important
5
Very strong more important
7
Extremely more important
9
Intermediate value
2,4,6,8
2. Rating the importance of the factors using the pair-wise comparison. We write the matrix
A to compare each criterion to the others.
1 / 2
 1

2 / 1
1
A
 ...
...

 n / 1 n / 2
.... 1 / n 

... 2 / n 
...
... 

...
1 
Find the eigenvector by normalized the pair-wise comparisons. We denote with vector s the sum
of the columns of the matrix and Divide each entry by the total of the column. Divide the total of row
by the total of number of row. Rate each factor relative to the others, base of degree of risk, for each
selected factor. So, doing the pair-wise comparison of the choices.
Normalize the pair-wise comparisons. Calculate index of consistence
max
is the eigenvalue calculated for all matrix of
n n
dimensions,
IC 
max  1
n 1
where
max   i  si , si are
i 1,4
the sums of each columns of the matrix by i.
Table 2 The IC and
n
IC
 max
1.
1
-
2
-
3
0.58
3.11
 max
4
0.9
4.27
limited values for all nxn matrix
5
1.12
5.44
6
1.24
6.62
7
1.32
7.79
8
1.41
8.89
9
1.45
10.1
10
1.49
11.3
Combined step 2 and 4 to obtain a relative rating for each choice.
3. Real data. Analytic Hierarchic Process.
37
This study is focused in the data base of Tirana Hospital, to evaluate the risk of getting cancer based
on these factors that are smoke, alcohol, obesity, family history. The data base is obtained during the
years 2010-2015, the mostly types of cancer are: throat cancer, lung cancer, breast cancer and prostate
cancer. We will construct the hierarchical structure, where the decision maker is considered cancer
risk, as the first level of the structure. In the second level are the factors as: smoke, alcohol, obesity,
family history. In the third level are cancer types as: throat, lung, breast, prostate. The hierarchical
structure is shown on Figure 1.
Fig 1. The hierarchical structure of this data set
Table 4. Comparison matrix of the factors in the 2nd level
Risk factors
Smoke
Alcohol
Smoke
Alcohol
Obesity
Family history
Column sum(
si)
1
1/2
1/4
1/5
1.95
2
1
1/2
1/6
3.66
Family
Obesity history
4
5
2
6
1
4
1/4
1
7.25
16
1

1/ 2
A( factors)  
1/ 4

 1/ 5
2
1
1/ 2
1/ 6
4
2
1
1/ 4
5

6
4

1
This sample of data has an information for n=150 patients. It contains four factors per person. We
have done a questionnaire to the doctors of this hospital in order to construct the pair-wise matrix,
based on the odds ratio of these main factors. Table 4, shows the risk factors pair-wise comparisons,
for the second level of the hierarchy.
Firstly, we calculate the eigenvalue
max
for this matrix, and the vector of relative weights as
follow:
.
1  4 1  2  4  5  2.5 2  4 1 1 2  6  1.56
2
3 
4
1 1
 1 4  0.84
4 2
4 
4

1 1
i
 1 4  0.3
5 6
i 1,4
 5.2
The normalized vector is:
 2.5 1.56 0.84 0.3 
 
,
,
,
   0.48, 0.3, 0.16, 0.06 
 5.2 5.2 5.2 5.2 
38
We denote with vector s  1.95,3.6, 7.25,16  the sum of the columns of the matrix.
max 

i 1,4
i
 si  0.48 1.95  0.3  3.6  0.16  7.25  0.06 16  4.1 .
The matrix is consistent because the eigen-value:
max  4.1 max   4.270 .
Now, we have to construct, for the third level, the pair-wise comparisons for each type of cancer
depending on the factors selected. Next we have to find for each table the eigenvector and then rank
the results by Table 5, Table 6, Table 7, and Table 8.
Table 5 Comparison of the cancer types related to smoke factor.
Smoke factor Lung
Prostate
Breast
Throat
Lung
1
3
1/5
1/9
Prostate
1/3
1
3
1/3
Breast
5
1/3
1
3
Throat
9
3
1/3
1
Column sum
15.33
7.33
4.53
4.44
Table 6 Comparison of the cancer types related to alcohol factor.
Alcohol factor Lung
Prostate
Breast
Throat
Lung
1
3
7
9
Prostate
1/3
1
5
7
Breast
1/7
1/5
1
3
Throat
1/9
1/7
1/3
1
Column sum
1.58
4.34
13.33
20
Table 7 Comparison of the cancer types related to obesity factor.
Obesity factor Lung
Prostate
Breast
Throat
Lung
1
3
1/4
2
Prostate
1/3
1
1/7
2
Breast
4
7
1
6
Throat
1/2
½
1/6
1
Column sum
5.83
11.5
1.4
11
Table 8 Comparison of the cancer types related to family history factor.
Family history Lung
Prostate
Breast
Throat
factor
Lung
1
4
6
7
Prostate
1/4
1
3
4
Breast
1/6
1/3
1
2
Throat
1/7
¼
1/2
1
Column sum
1.55
5.58
10.5
14
Table 9 Local weights for every table above.
39
max
Tab 5
3.96
Tab 6
4.26
Tab 7
3.8
Tab 8
3.9
Tab 4
4.1
(vector )
(lung ) =(0.11,0.16, 0.33, 0.38)
( prostate) =(0.58,0.29,0.085, 0.04)
(breast ) =(0.19, 0.096, 0.63, 0.078)
(throat ) =(0.61, 0.22, 0.09, 0.06)
( factor ) =(0.48, 0.3, 0.16, 0.06 )
 lun  pros bre thro  factor
 0.11 0.58 0.19

 0.16 0.29 0.096
 0.33 0.085 0.63

 0.38 0.04 0.078
0.61 

0.22 
0.09 

0.06 
 0.48 


0.3 


 0.16 


 0.06 
result
 0.286 
=

 0.185 
 0.280 


 0.207 
Ranking
1
 
 4
 2
 
 3
This gives a final priority of 0.285 for lung cancer, 0.185 for prostate cancer, 0.280 for breast cancer
and 0.207 for throat cancer.
Conclusion
We have classified the cancer types, by the factors included in level 2, using the pair-wise comparison
method of AHP. The AHP methodology applied to the data collected from Tirana Hospital, shows
that lung cancer is mostly spread in humans with the highest risk, and prostate cancer is the least
spread in humans, with the smallest risk. The model chosen is a method for helping the decision
maker to the problem that has a limited number of choices, with a limited number of factors at each
choice.
Acknowledgments
This work is part of my doctoral study research at the University of Tirana, Faculty of Natural
Sciences, Department of Applied Mathematics. We acknowledge financial support from the
University “Aleksander Moisiu” Durres.
References
1. R.W.Saaty , The Analytic Hierarchy Process-What is and how it is used, math modeling, Vol 9,
No 3,4,5,1987, 161-176.
2. L.Abdullah, I.Taib, R.Salleh, Public perceptions of cancer risk using Analytic Hierarchy Process
, Journal of applied sciences, volume 12, nr 9, 2009, p 2319-2324.
3. Calle, E.E. and M.J.Thun, 2004. Obesity and cancer , 23; 6365-6378.
4. Liana Najib, Lazim Abdullah, Ilyani Abdullah, Zabidin Salleh, Weights of road accidents causes
using Analytic Hierarchy Process, ARPN Journal of Science and Technology, vol 2, nr 2, 39-44,
2012.
5. Dyer, J.S, 1990b, A clarification of remarks on the Analytic Hierarchy Process. Manage. Sci 36,
274-276.
40
6. Kim,S.E, E.J.Perez-Stable, S. Wong, S. Gregorich, G.F.Sawaya, J.M.E Walshand C.P.Kaplan
2008, Association between cancer risk perception and screening behavior among diverse woman.
Arch.Int.Med, 168, 30-700.
7. Kamal. Al-Subhi Al-Harbi, Application of the AHP in project management, International Journal
of Project Management19, 2001.
41
Implementation of E-education in Africa via Space
Networks
Dimov Stojce Ilcev
Durban University of Technology (DUT), Durban, RSA,
ilcev@dut.ac.za
Abstract. In this paper is introduced an advanced E-education provision in remote dispersed
communities, such as rural, remote, mobile, agriculture, mining, construction, surveying, military,
tourism and so on, based on the specific needs and requirements, implies significant broadband
connectivity requirements, timely and quality-assured content delivery of service, applications and
interactivity. The E-education solutions of distance learning and training for remote and rural areas,
which are out of range of terrestrial and short distance wireless telecommunications facilities, cannot
provide the broadband access without space-enabled communication infrastructures, such as satellite
constellations and Stratospheric Platform Systems (SPS) or High Altitude Platforms (HAP). The
paper also discusses the integration challenges that are presented by combining use of the space
solutions for implementation Education and learning in urban, rural and mobile environments. The
configuration of in-house design and development of Space Segment, installation of the scale-down
Digital Video Broadcasting-Return Channel via Satellite (DVB-RCS) Hub as a Gateway, Ground
Network and Interactive VSAT, known as Fixed Interactive Terminals (FIT), for E-education,
distance learning and staff training initiative in Africa are described.
Keywords: DVB-RCS, DVB-S/S2, VDVoIP, Space and Ground Segment, Hub
1. Introduction
In the past are designed many proposals employing ICT and software support to provide E-education
in remote environments, but these solutions couldn’t work at all without implementing two-way
space connectivity via Geostationary Earth Orbit (GEO), Medium Earth Orbit (MEO), Low Earth
Orbit (LEO), SPS or Unmanned Aerial Vehicles (UAV), illustrated in Figure 1. Usually the best
satellite solutions for E-education are GEO constellations deploying DVB-RCS standards and SPS
with new proposed Digital Video Broadcasting-Return Channel via Platforms (DVB-RCP) technique
and technologies. In addition, it will be necessary to examine a set of services and infrastructures that
will realize E-education and distance learning systems for rural schools and corporate organizations
and to assess the targeted users’ interest in such specific applications.
42
Fig. 1. Space Constellations in Function of E-education – Courtesy of Manual: by Ilcev [2]
The growing demand for satellite or platform communications bandwidth over all wireless media has
motivated the introduction of digital solutions that provide bi-directional services at greater than 300
MB/s known as DVB-RCS or RCP. Current DVB applications may demand rapid E-education
installations at fixed locations of urban, suburban and rural environments, as well as mobile
platforms, such as buses or vehicles for E-learning.
In such as way, more than a basic introduction, the DVB project presents the specific approaches to
design and select an appropriate broadband satellite constellation deploying C, Ku and Ka-bands.
Then, it needs to configure ground elements, evaluate sources of antennas technology, Radio
Frequency (RF) electronics equipment, services and solutions. This new technique can take into
consideration all applications to commercial and government users as well as compliance with
International Radio Regulations of International Telecommunication Union (ITU) [1], [2], [3].
2. Development of E-education via Stratospheric Platform Systems (SPS)
The SPS technologies will provide implementation of DVB-RCP access via airship for fixed and
remote broadband connecting urban, rural and mobile platforms anywhere and in anytime. In fact,
the development of the DVB-RCP infrastructure for E-education will need design of special airship
that will act as geostationary very low orbit satellite at about 20 to 50 km in stratosphere, which
network is shown in Figure 2. The SPS network contains space and ground segment and one airship
can cover up to 500 km in radius.
43
Fig. 2. E-education Solution via SPS Airship – Courtesy of Manual: by Ilcev [2]
Space segment includes one or several airship platforms connected via Inter Platform Links (IPL)
using laser or optical links. Thus, deploying few airships with overlapping coverage will be covered
a large land masses over one country. The ground segment consists Hub (Gateway) station interfaced
to the Terrestrial Telecommunication Network (TTN), Internet Service Providers (ISP), Integrated
Service Digital Network (ISDN), Broadband ISDN (B-ISDN), Asynchronous Transfer Mode
(ATM), Universal Mobile Telecommunications System (UMTS), General Packet Radio Services
(GPRS), G4 or G5 Cellular Networks and Broadcast E-learning Centre (BEC). On another side are
connected users segment of rural schools to the BEC site via one or few airship platforms. Thus,
educators in BEC can provide lessons of many subjects in the real time via Videoconference (VC) to
the rural classrooms equipped with DVB Fixed Interactive Terminal (FIT) or Very Small Aperture
Terminals (VSAT). On the roof of rural school or mast will be fixed DVB antenna as an Outdoor
Unit (ODU) and inside of school will be installed DVB router or Indoor Unit (IDU). Router is able
to connect up to 100 PC terminals in Local Area Network (LAN) by the cable lines or via Worldwide
Interoperability for Microwave Access (WiMAX) and Wireless Fidelity (Wi-Fi) by the wireless
links. The video lessons coming from BEC in rural schools have to be viewed via PC or common big
VC screen and educator for each subject will be able to send soft copies of books and other learning
materials. Pupils in rural schools can watch IPTV (IP Television) as well.
44
Fig. 3. DVB Space Segment Coverages – Courtesy of Brochures: by Eutelsat/Arabsat [4]
However, the schools offices, rural households, Small offices/Home offices (SoHo), Small/Medium
Enterprises (SME), rural Internet kiosks and other sites will be connected to Internet, VC, IPTV,
VoIP telephones, VDVoIP units and G4Fax machines. Managers of schools may be in touch in any
time with their supervisors in Education Department and educators in BEC. Thus, as alternative can
be used Portable E-learning VSAT.
From the geometrical point of view, the SPS airship terminals would enable very reliable rural
communications that take much advantage of the best features of both terrestrial and satellite
multipurpose communications. In addition, the system could bring advantages of its own, not
available in current systems. The most important advantages of employing SPS are high elevation
angles, broad coverage, low propagation delay, extremely low-cost operation, easy and incremental
deployment and ability to move around the area in emergency situation.
The SPS airship can be launched using a specified volume of helium gas separated from the air to
maintain its shape. As the platform rises the helium expands and at the proper altitude displaces all
of the air within the airship. Once it is in the stratosphere the airship is remotely controlled and moved
into determined position. The launch of SPS into position is much simpler than putting a satellite into
any orbit. After careful preparation in the hangers, the airship is launched in 4 Ascent phases through
the troposphere and Interface location point in the stratosphere and finally, it shifts to the stationkeeping position. The recovery phase goes in the opposite direction, namely, the airship is slowly
moved from the station-keeping position towards the Interface point and from there descends down
to the ground in 4 descent phases.
The SPS airships do not interfere aircrafts flights, because they are located over 10 Km, airship itself
leverages new Lighter-Than-Air (LTA) technology being made of very high strength and lightweight
materials. Airship is unmanned and solar powered platform, accompanied by advanced propulsion
systems that maintain proper positioning, and it is also equipped with autonomous navigation, radio
controlled command and communications payload stabilization systems. A combination of solar
cells, batteries and fuel cells will power the SPS during its five-year planned deployment.
45
Fig. 4. DVB-RCS Ground Segment – Courtesy of Brochures by ViaSat/Cobham/Hughes [6]
Thus, SPS also incorporates telemetry to remotely transmit data (TT&C) and redundant systems to
serve as back-up measures, then features specific concept to provide the platform with very high level
of availability, reliability and safety. The SPS is being designed to hold approximately 1,000 kg of
communications payload capable of supplying fixed and mobile broadband, narrowband and wireless
backbone services to approximately 3 million subscribers.
Immature SPS airships, their stabilization system and onboard antenna technology are challenging to
be investigated, to avoid some inherent limitations belongs to the traditional systems and to provide
backbone to cellular networks. Those are in the sense of a huge number of base stations required by
the terrestrial system, limitation of the minimum cell size on the ground involved in GEO satellite
system, and suffer from handover problem faced by LEO/MEO satellite system. With these great
advantages, the ITU has allocated the spectrum to this system at 2 GHz for 3G mobile systems, 48/47
GHz for the usage worldwide, and 31/28 GHz band is allocated for usage in many countries [1], [2],
[3], [4], [5].
3. Development of E-education via Satellites
The first generation of satellite standard of DVB-S known as an DVB-RCS, about 15 years ago,
quickly became around the globe one of the key solution in almost every new advanced satellite
communication projects for broadcast, broadband and multimedia interactive applications including
high-speed Internet, IPPC and IPTV. The second generation of DVB-S2 CCM (Constant Coding
Modulation) standard few years ago was presented as a new more cost effective, efficient, reliable,
secure and functional solution. The DVB-S2 CCM is recently upgraded by the most technical and
cost effective mode of DVB-S2 ACM (Adaptive Coded Modulation) standard as a forward and
reverse compatible.
46
Fig. 5. E-education Solution via GEO Spacecraft – Courtesy of Book: by Ilcev [5]
The new DVB-S3 Third Generation of DVB-S Standard was proposed by the Israeli company
NovelSat as more efficient successor to the DVB-S2 transmission system. After launching its new
DVB-S3 (NS3) modulation technology in April 2011, as more efficient than DVB-S2, start-up
NovelSat Company has already established 32 live satellite trials. The system firstly was deployed
in a series of VSAT terminals and targeted as a replacement technology for DVB-S2 delivering VDV,
Internet and IPTV from the sky to consumer homes and mobiles in remote areas.
For establishment DVB-RCS network in Northern and Southern Africa can be employed existing
satellite operators providing the C, Ku and Ka-band Satellite Constellations suitable for DVB-RCS
S and S2 scenario, which beam coverages are illustrated in Figure 3 (Left) of Eutelsat and in (Right)
of Arabsat. There are many other DVB satellite operators such as Intelsat, SES-NewSkies,
PanAmSat, Telesat, Inmarsat and others providing regional or global and spot beam coverage via
GEO satellites. Presently in Africa are operational for regional coverage two satellites only: the
Egyptian multipurpose spacecraft Nilesat and Nigerian NigComSat.
The British company Inmarsat awarded a contract to the US Boeing in August 2010 to design, build
and supply three Inmarsat-5 satellites as part of an 752M £ (1.2B $) wireless broadband network
known as Global Xpress.
47
Fig. 6. E-education in Rural Schools via DVB-RCS Solution – Courtesy of Brochure: by
NanoTronix [7]
This network will operate at Ka-band frequencies ranging between 18GHz and 31GHz, which
however possess more signal quality issues, compared to that of old Ku-band (12GHz-18GHz)
frequency range. Africa and Middle East have 69 countries, so dividing 1.2B $ with 69 will be
participation of about 17M $ per each country in the region to build 3 and 5.5M $ for 1 multipurpose
GEO spacecraft.
The DVB-RCS Hubs are turnkey cost effective systems that can be installed in days to enable a wide
range of government, corporate and private public network topologies with satellite interactive
terminals including rural and learning applications, which two versions are shown in Figure 4A:
ViaSat (Left) and Hughes (Right). Depending of type, but the biggest Hub terminals are able to
support up to 80,000 VSAT or router terminals in urban, rural and remote environments, which two
versions are shown in Figure 4B: ViaSat (Left) and Hughes (Right), and in Figure 4B (Above) is
shown Cobham Patriot DVB-RCS antenna.
Thus, the DVB-RCS VSAT network has been designed to minimize the cost of scaling a broadcast,
broadband, Internet and multimedia access between Hub terminals and hundred or thousands
simultaneously logged-on FIT units in rural or remote areas providing E-education via GEO Satellite,
shown in Figure 5. This E-education solution has almost the same as SPS infrastructure and solutions
shown in Figure 2. In addition, the DVB-RCS VSAT network is providing Mobile E-learning
solutions in special terrain vehicles and in school buses.
Each E-education vehicle integrates complete VSAT indoor (IDU) routers and on top of vehicle are
installed VSAT outdoor (ODU) reflector antennas at C (4-8 GHz), Ku (12-18 GHz) or Ka-band (2740 GHz). The VSAT equipment is connecting several PC configurations in LAN inside vehicles or
via cables inside schools for interactive videoconference with Broadcast E-learning Centre (BES) in
urban area. Educator is providing live lecture of remote education of any subject for each rural school,
portable or mobile E-learning solution.
In Figure 6 is shown block diagram of typical rural E-education solution that connects urban and
rural areas via DVB-RCS Hub, Radio Frequency (RS) antenna and C/Ku/Ka-band GEO Spacecraft.
On the urban side is ISP connected to Router, Backbone Network controlled by the Network
48
Management System (NMS) and via Internet Protocol (IP) is linking Hub. The rural area contains
the number of remote schools, which are equipped with VSAT routers, antenna and many PC
terminals in LAN, WiMAX or WiFi networks. In every school pupils are watching live lecture at
their PC monitors or videoconference screens. At the end of live lecture for each subject classroom
teacher is asking pupils for any misunderstandings and is transferring transfer all questions to the
educator in BEC studio for interactive discussion [2], [5], [7], [8].
Conclusion
The new DVB-S2 is designed to minimize overall system costs for service providers and system
operators. Thus, this network offers the lowest costs on the market today for multiple access systems
managing VoIP, VDVoIP, IPPC and IPTV multimedia, broadcast and broadband contents. The Eeducation solution throughout DVB-RCS will improve education and learning facilities in rural and
urban areas and will help for better knowledge, information system and service delivery. These
solutions via DVB-RCS will also help all schools, pupils and teacher to be better managed,
controlled, inspected and educated countrywide. In all events, the implementation of DVB-RCS
architecture will improve communication and Internet facilities in rural and remote environments for
government and private corporation including emergency, disaster, security, education, health
solutions and additional E-solutions.
References
1. Maral G., “VSAT Networks”, John Wiley, Chichester, 2003.
2. Ilcev D. S. “Global Mobile CNS”, DUT, Durban, 2011.
3. Zavala A.A. & Others, “High-Altitude Platforms for Wireless Communications”, Wiley, 2008.
4. WebPages: “GEO Satellite Operators: Eutelsat and Arabsat, 2008.
5. Ilcev D. S. “Global Aeronautical Communications, Navigation and Surveillance (CNS)”, AIAA,
Reston, 2013.
6. Web, “ViaSat [www.viasat.com], Cobham [www.cobham.com, Hughes [www.hughes.com]”.
7. NanoTronix, “Two-way VSAT for E-learning”, Seoul, 2011 [www.nano-tronix.com].
8. Ilcev D. S., “Stratospheric Communication Platforms (SCP) as an Alternative for Space Program“,
AEAT Journal, Emerald, Bingley, 2011.
49
Introduction to Stratospheric Communication Platforms
(SCP)
Dimov Stojce Ilcev
Durban University of Technology (DUT), Durban, RSA,
ilcev@dut.ac.za
Abstract. In this paper are introduced the modern airship techniques and technologies as cost
effective solutions of Stratospheric Communication Platforms (SCP). The launch or putting in
position the airship is not critical point such as launch of satellite and controlling support services in
the creation of space-based communication technology and the most expensive phase of the total
system cost. Therefore, with few cost effective remote controlled and solar powered airships can be
covered some region or country including urban, suburban and rural areas, mobile, farms and other
environments with low density of population. The airship SCP network offers better solutions than
cellular radio systems, with greater speed of transmission than even optical modes, roaming will be
enhanced without severe shadowing or obstacle problems and disturbances inside of buildings and
service will cost less. The SPS mission system is more autonomous and discrete, can be integrated
with current satellite and cellular systems, and will be the best solution for rural, mobile
transportation and military applications. The SCP airship can be seen well from all positions inside
coverage area, because they are overlapping the total coverage and because of elevation angle. In any
circumstances mountains, buildings and even trees cannot cause obstructions like to cellular network.
For these reasons, there is currently a revival of interest for SCP constellations and application types
of various system concepts are being studied.
Keywords: SCP, TAO, Airship, Ground Segment
1. Introduction
Despite of advances in Wireless and Terrestrial Telecommunications Technology, almost three
billion of world populations are living in rural locations are still without any telephone facilities.
However, except already developed Fixed and Mobile Satellite Communications Networks, SCP
Systems are the newest space technique with top digital transmission technologies for fixed and all
mobile commercial and military applications, which include remote and rural solutions. These
systems employs unmanned or manned and on solar or fuel energy aircraft and airships carrying
onboard payloads with transponders and antennas, providing new technique known as Voice, Data
and Video over IP (VDVoIP) service.
The SCP networks are the newest space technique with top technologies for fixed and mobile
applications, including military solutions. These systems are using unmanned or manned aircraft and
on solar or fuel energy airships and carrying payloads with transponders and antenna systems. With
a few very cheap remote controlled and solar powered airships as a better solution, a territory can be
covered of some region or country including urban, suburban and rural areas, farms and other
environments with a low density of population. However, today there are developed four general
telecommunications architectures, which can be used to deliver broadband wireless local loop service
to consumers. Two of these architectures are Geostationary Earth Orbit (GEO) and Non-GEO
satellite systems and the other two are terrestrial rooftop cellular-like millimeter wave repeaters and
stratospheric relay platforms.
50
The SCP network offers better solutions than all cellular and wireless systems, with greater speed of
transmission than even optical modes, roaming will be better, without shadowing problems and
disturbances inside buildings and service will cost less. The SCP mission can be integrated with
current satellite and cellular systems; the system is more autonomous and discrete and will be the
best for military and all mobile applications. For instance, the Halo Broadband GSPS Millimetre
Wavelength (MMW) Network of the Angel Company provides data densities nearly one thousand
times higher than proposed satellites, shown on Table 1, while having round trip time delays
appropriate for interactive broadband services. Whereas, the delays through satellite network nodes,
even through LEO satellite nodes, are too long for many interactive applications, delays are 25 or
1,000 times longer for LEO or GEO then for Halo Networks, respectively. In fact, the Halo
comparison parameters are similar to a variety of metropolitan environment spectrum used bands of
the Local Multipoint Distribution Service (LMDS) band near 28 GHz [1, 2, 3].
Table 1. Comparison of Data Density and Signal Delays [2]
Node Type
LMDS
Halo
LEO (Broadband)
GEO
Node Data Density
Min
Max
(Mb/s/km2)
(Mb/s/km2)
3
30
2
20
0.002
0.02
0.0005
0.02
Round Trip Delay
Min
Max
(millisec)
(milieu)
0.003
0.060
0.10
0.35
2.50
7.50
200
240
2. Airships SCP
The new airship projects offer cost-effective systems for SCP by using special unmanned and nonfuel solar cell powered balloons with an estimated endurance of several months. In comparison of
aircraft and airship systems it is difficult now to say which one will be better for the future reliable
SCP. Thus, there are several airships such as: SkyStation Global Network, SkyLARK Network,
StratCon (StratoSat) Global Network, TAO Network, etc.
Fig. 1. TAO Airship with Main Components - Courtesy of Webpage: by TAO [3]
51
3. TAO (SkyNet) Network
A Research and Development program (R&D) on a SCP airship system is in progress since April
1998. The final goal of this project is to realize the SCP airship platform system, being capable of an
acceptable long-duration station-keeping flight at a stratospheric altitude of about 20 km. The
achievements will enable advanced wireless fixed and mobile communications, digital direct and
relay broadcasting, modern broadband and multimedia transmission, high speed Internet, highresolution observations and monitoring of the remote, rural and global environment. This advanced
SCP program is recently promoted in collaboration with the Communications Research Laboratory
of Japan (CRL), National Space Development Agency of Japan (NASDA) and Japan Marine Science
and Technology Centre (JAMSTEC), including the Telecommunications Advancement Organization
(TAO) of Japan.
3.1 Airship Platform System Description
The stratospheric platform is an unmanned airship kept at a stratospheric altitude of about 20 - 25 km
for broadcast and multimedia communications and Earth observation purposes, illustrated in Figure
1. The SCP airship is equipped with corresponding communications payload, observation sensors
and other necessary flight equipment. Thus, with the aim of quickly developing an SCP platform has
great potential; so many research institutions in Japan began conducting the relevant research work
in 1998. The SCP system is designed similar to a satellite space segment as a relay station to receive
signals from ground stations using feeder links and to retransmit them to subscribers using service
links.
Therefore, an airship platform like a satellite is carrying a payload with corresponding transponders
and antenna system. At any rate, the launch of SCP into position is much simpler than putting a
satellite into any orbit. After careful preparation in the hanger space, the airship is launched in 4
Ascent phases through the troposphere and Interface location point in the stratosphere and finally, it
shifts to the station-keeping position. The recovery phase goes in the opposite direction, namely, the
airship is slowly moved from the station-keeping position towards the Interface point and from there
descends down to the ground in 4 descent phases. The airship construction has a semi-rigid hull of
ellipsoidal shape, with an overall length of about 200 m. Platform is composed of an air-pressurized
hull for maintaining a fixed contour and internal special bags filled with the buoyant helium gas. Two
air ballonets are installed inside the hull to keep the airship at a required attitude. For a load balance
to the lifting force, catenary curtains are connected to the lower rigid platform’s keel and are directly
attached to the envelope. Propulsive propellers are mounted on both the stem and stern of the airship
and tail fins are installed on the rear end of the hull. A solar photovoltaic power system of solar cells
and Regenerative Fuel Cells (RFC) is provided to supply a day and night cycle of electricity for
airship propulsion. The length of an airship in general is about 250 m and 60 m diameter. This is
about 4 times as long as Jumbo jet passenger airplanes and so its weight is about 32 tons. However,
50% of the weight corresponds to those of structures and membrane materials. Hence, solar arrays
and fuel batteries, related to the electric power subsystem, are also heavy. And the weight of mission
equipment is supposed to be about 1 ton. The necessary condition for an airship to float at a certain
altitude is that the gravity and buoyancy forces, which are exerted on the airship, are in a state of
equilibrium. When the shape and volume of the airship are supposed to be constant, unlike a balloon,
the buoyant force at an altitude of 20 km becomes about 1/15 that at sea level. Accordingly, a
buoyancy of 15 times as much is necessary for equilibrium. Therefore, in order to float a SCP in the
stratosphere, it is necessary to make the weight of the airship light and to make the buoyancy as large
as possible. Inside the airship there are several internal bags filled with He gas to obtain enough
buoyancy [2], [3], [4], [5].
3.2 Outline of the Tracking and Control System
52
In order to operate unmanned SCP airship safely, it is necessary to construct a tracking and control
system and establish operational technique on board the platform and on the ground. Based on SCP
technologies, appropriate countermeasures can be taken in time regarding observation and prediction
of weather situations and monitoring of operational conditions of onboard equipment, and even
regarding safety of dangerous atmospheric phenomena or abnormal performances of onboard
equipment. At this point, the TAO airship system has to develop adequate TT&C solutions on board
the platform and on the ground as well.
During launch airships can be affected strongly by wind, therefore, when the preliminary decision
for launching or recovering of an airship is to be made, it is necessary to predict the weather data,
especially wind direction and speed, in advance and estimate whether:
1) The airship deviates from the area, within which the tracking and control system works effectively,
and
2) The launch of SCP airship or recovery can be conducted safely.
Based on this proper estimation, a final decision to launch or recover is made safely. After the last
checks, the airship is released. It starts to ascend due to the effects of the buoyancy. Near the
tropopause, which is the layer between the troposphere and stratosphere, it continues to ascend, being
drifted by the jet stream. Finally, the airship arrives in transfer position at an altitude of about 20 km.
After this operation, the airship is moved to the geo station-keeping position and the mission
operation is started. Once an airship is launched, it can be used for a maximum of three years.
Namely, an airship is periodically, about every three years, recovered and the He gas and onboard
equipment condition is checked. Moreover, after these routine checks, it is launched again. The life
of an airship is supposed to be more ten and even twenty years.
3.3 SCP Network Coverage
The main question of the future system is how many airship platforms are necessary to cover all
particular territory or country and can this system be global? In fact, a 15 Stratospheric Platform
arrangement is necessary to cover all the territory of the Japanese islands for communications and
broadcasting systems, under the condition of 22 km airship altitude with a minimum elevation angle
of 10o. A single airship can cover a certain service area independently, so that, for example, the
service can be started from an area with a large density of population with the number gradually
increased. This possibility of flexible business development is one of the merits of SCP systems. The
service area enables that one airship can cover generally depends on certain available numbers of
ground Transmitter (Tx), Receiver (Rx), two-way direction antennas, methods of modulation and
transmission and many other factors. Otherwise, the final intention of this project is to offer services
to other regions and certain countries and if economical and technical evaluations are correct, it will
provide global coverage. The concept of the system is very advanced in comparison with similar
projects and has almost no disadvantages.
3.4 Ground Segment Features
On board the airship there is mission equipment to provide Multimedia and Broadcasting for Fixed
and Mobile communications including an Earth observation and Disaster Monitoring System. So,
airship is expected to have the following features:
53
Fig. 2. Fixed and Mobile Communication Network via SCP- Courtesy of Book: by Ilcev [2]
1) Broadband communications and broadcasting are possible with small-sized and very low-power
terminals, because of much shorter propagation distances compared with satellite systems.
2) High-quality, reliable and very cost-effective communications and broadcasting are possible with
a smaller number of ground stations, due to significantly better line-of-sight conditions, less waveblocking and multi-path effects compared with existing ground systems.
Meanwhile, compared with satellite systems, the propagation distance is shorter by about 1/1800.
Consequently, as Electro Magnetic (EM) radiation propagation losses and delay distortions become
much smaller, broadband communications and broadcasting are possible with smaller sized and
lower power fixed and mobile terminals.
3) By establishing interplatform links, high-speed communications and broadcasting networks,
comparable to optical fiber systems will be possible including novel communications.
4) Optimum communication configurations links are possible owing to the flexible operations of
airship SCP systems, which can enable expansion to a world communications system.
A complete inter-platform links, fixed and mobile broadband multimedia service program with all
applications is presented in Figure 2, [2], [4], [6], [7].
54
Fig. 3. Functional Diagram of SCP Subscriber Equipment - Courtesy of Book: by Ilcev [2]
Digital TV Broadcasting by SCP will use about 15 CSP to cover all of Japan, where over 10,000
stations are necessary at present. This system will be an advanced communications complement to
terrestrial systems at very low cost. Thus, access to for wireless communications systems will be
more cost-effective than optical fiber systems. The TAO system will enable much advanced, reliable
and better communications for developing countries and will promote spreading (diffusion) in these
countries. Emergency communications systems will retain communications links in an emergency
by moving into position over the area. In addition, various remote-sensing services will be available
for radio wave observations, aerial photographs, Earth and meteorological observations and so on.
The CSP System is designed for fixed and mobile multimedia two-way communication system. The
ground segment consists of Ground Earth Stations (GES) or Gateways and fixed, semi fixed and
mobile terminals, with onboard corresponding auto tracking and focusing antenna systems for all
applications, respectively. In fact, fixed ground terminals can be a self-contained portable or office
PC configuration with modem, or as an integrated part of an advanced LAN or WAN, laptop, video,
fixed telephone set in office or public and mobile or cellular phone equipment.
Mobile systems for Communication, Navigation and Surveillance (CNS) will offer maritime, land,
aeronautical and personal applications including GNSS access. At this point, mobile user terminals
can be PC/laptop portable or fixed configurations interfaced to the SCP transceiver with adequate
antennas or self-contained mobile or portable/in vehicle transceiver units with mobile auto tracking
antenna and personal handheld terminals with built-in antenna.
In Figure 3 is shown block diagram of prototype for fixed and mobile SCP equipment and their
services.
Maritime SCP equipment and system will provide two-way commercial, distress, emergency and
safety communications for ships, fishing and other boats integrated with GEO and GNSS systems.
In the framework of this service, there will be additional activities like buoy and lighthouse control,
marine pollution, investigation, warnings and Search and Research (SAR) missions.
Land mobile equipment and systems will provide services for all kinds of vehicles like trains, buses,
trucks and cars, including personal mobile units, cellular service and emergency communications for
natural disasters, which can be enhanced with equipment for tracking and navigation facilities. The
SCP has to substitute or integrate current cellular systems.
Aeronautical mobile equipment and systems will provide commercial, safety and distress service for
all kinds of aircraft integrated with GEO, GPS and other Global Navigation Satellite Systems (GNSS)
to provide CNS service.
55
The broadcasting system using the SCP airship constellation will provide: 1) digital broadcasting; 2)
complementary terrestrial digital broadcasting to fixed/mobile stations; 3) terrestrial rebroadcasting
programs; 4) relay broadcasting of HDTV including radio programs; 5) movable broadcasting on
demand, using mobile equipped stations; 6) broadcasting for limited suburban regions, isolated
islands, rural and remote places; 7) field pickup from the SCP; and 8) emergency news and
observations [2], [3[, [4], [8].
3.5 Electric Power and Motion of Airships
Unmanned SCP airships are maintained in the stratospheric zone constantly at an altitude of 20 to 25
km above commercial flights, where the winds, flow and other meteorological conditions are calm.
The mean temperature is –60 to –50o C and the atmospheric pressure is about 50 hPa. The altitude of
20 km is about 60 times higher than the Tokyo TV Tower, so better Lines-of-Sight (LoS) are
obtained. However, multi-path effects will be significantly reduced in most areas because of higher
elevation angles. There are no clouds in the stratosphere, so perfectly clean solar energy can be used
without atmospheric pollution. The concept of SCP airship launch and recovery features an upwind
orientation of the airship heading, assuming limited wind speed at the ground surface for takeoff and
landing. In additional, for ascents and descents, the airship heading and position should be kept
always in an upwind direction, while non-powered flights should require only the buoyant lift into
the jet stream.
Stratospheric platforms do not use any fuel for motion, so the only necessary energy is supplied by
electric power, obtained from solar cells. The wide part in the graphic of an airship’s topside is the
solar cell array. Solar cells can supply clean energy without CO2 gas generation, so that it is said to
be kind to the Earth and is an ideal form of energy. Only electric power of about 200 kW is necessary
for stratospheric platforms and it is used for the following reasons:
1) To rotate propulsive propellers and supply electric power to the various positioning and attitude
controlling systems for station keeping.
2) To operate the on-board payload for mission of communication, broadband and Earth observation
systems are necessary about 10 kW.
3) To charge the fuel battery in the daytime and to use it at night, when no solar electric power is
obtained.
56
The position of an airship is as stationary same as a GEO satellite. Namely, as the airship is used as a
platform for communications and broadcasting, it is necessary for it to be almost stationary, like a
broadcasting satellite.
Meanwhile, the position of the airship cannot be permanently stationary because of affects from
different parameters, which have an influence on the moving range. At this point, it is necessary to
provide permanent control of platform moving and when necessary to correct it’s position using
station-keeping correction motors with forward and rear propellers.
Consequently, as the maximum wind speed is sometimes over 20 m/sec at altitudes of 20 to 22 km, it
will be very difficult for the airship to be controlled as strictly as geostationary satellites. The moving
range of a stratospheric platform is very slow and in such a way requirements of station keeping are
within a radius of 1 km in the horizontal plane and +/–1km in the perpendicular direction and under
this condition the design and fabrication of onboard equipment is considered.
A receiving antenna on the ground is set with the bore sight (the axis of the antenna beam) to the
airship, so that the maximum receiving power can be obtained. If the position of the airship is changed,
the receiving power decreases.
The airship is controlled so as to keep the position as stable as possible. For example, for a 20 cm
diameter antenna, a position change of 600 m induces 3 dB or 50% power loss. This tendency depends
upon the antenna size and operational frequency. In this instance, the larger an antenna radius or the
higher the frequency, the receiving power becomes less and vice versa. Accordingly, to satisfactorily
solve these receiving problems, an automatically pointing corresponding adjustable antenna is being
considered and in this way, it is necessary to develop a low-cost and small-sized antenna with
automatic tracking system [2], [3], [4].
Conclusion
There are several unique attributes that allow modern SCP constellation, either aircraft or airships, to
offer low-cost broad array service:
1) The SCP infrastructures do not require launch vehicles, they can move under their own power
throughout the world or remain stationary, and they can be brought easily down to Earth surface after
several months of mission, refurbished and re-deployed without service interruption.
2) With only several SCP airships units is possible to provide regional or local countrywide coverage,
while with infrastructure of interplatform optical links between nearby SCP airships is possible to
enlarge coverage and enhance service.
3) Every SCP is independent and autonomous for dedicated service area, so once a platform is in right
position, it can immediately begin delivering service to its coverage area without the need to deploy
a global constellation of platforms to operate. This attribute is very important for rural, mobile and
military applications.
4) The altitude enables the SCP system to provide a higher frequency reuse between nearby platform
and higher capacity than other wireless systems.
5) Located in the stratosphere 20 to 25 km above the Earth, each SCP is acting as the very low orbiting
satellite, providing high density, capacity and speed service with low power requirements and no
latency to entire urban and rural areas.
6) The inexpensive SCP and Gateway stations make it the lowest cost wireless infrastructure per all
subscribers conceived to date. The flexibility, reliability, enhanced capabilities and its low-cost will
revolutionize telecommunications.
7) Joint venture private companies, government and other authorities located in each country will
control the SCP configurations serving their region to ensure the best service offerings tailored to the
local market.
8) All types of SCP missions are environmentally very friendly, because most of them are powered
by solar technology and non-polluting fuel cells.
9) The SCP airship provides all subscribers with short paths through the space and unobstructed free
of shadowing LOS. With very small antenna, transceivers and low power requirements, the SCP
system allows for a wide variety of fixed, mobile and portable user terminals to meet almost any
service need including distress, commercial and military applications.
57
References
1. Aragón-Zavala A., “High Altitude Platforms for Wireless Communications”, Wiley, Chichester,
2008.
2. Ilcev D. S. “Global Mobile CNS”, DUT, Durban, 2011.
3. TAO, “Stratospheric Platforms”, Tokyo, 2006.
4. Grace D. and Mohorcic M., “Broadband Communications via High-Altitude Platforms”, Wily,
Chichester, 2011.
5. Tozer T.C. and Others, “High-altitude Platforms for Wireless Communications”, Electronics and
Communication Engineering Journal, 2001.
6. Miura R. and Others, “Experiments on IMT-2000 Using Unmanned Solar Powered Aircraft at an
Altitude of 20 km”, IEEE Transactions on Vehicular Technology, Vol. 54, No. 4, July 2005.
7. Ilcev D. S., “Stratospheric Communication Platforms (SCP) as an Alternative for Space
Program”, AEAT Journal, Emerald, Bingley, 2011.
8. Antonini M. and Others, “Stratospheric Relay: Potentialities of New Satellite-high Altitude
Platforms Integrated Scenarios”, IEEEAC, 2003.
58
The Importance of Big Data Analytics
Eljona Proko
Computer Science Department, Vlora University
Vlore, Albania
elzavalani@gmail.com
Abstract. Identified as the tendency of IT, Big Data gained global attention. Advances in data
analytics are changing the way businesses compete, enabling them to make faster and better decisions
based on real-time analysis. Big Data introduces a new set of challenges. Three characteristics define
Big Data: volume, variety, and velocity. Big Data requires tools and methods that can be applied to
analyze and extract patterns from large-scale data. Companies generate enormous volumes of polystructured data from Web, social network posts, sensors, mobile devices, emails, and many other
sources. Companies need a cost-effective, massively scalable solution for capturing, storing, and
analyzing this data. They also need to be able to integrate their Big Data into their real-time analytics
environment to maximize business value. Big Data was seen as a mean to manage and to reduce the
costs of data management. In this paper we discuss definition, characteristics, challenges, techniques
for analyzing Big Data and the importance of Big Data. The main importance of Big Data consists in
the potential to improve efficiency in the context of use a large volume of data, of different type. Big
Data enables companies to create new products and services, and invent entirely new business models.
A case study will be present to demonstrate the potential of Big Data Analytics: Oracle Application
in Social Insurance. We conclude this paper that through better analysis of the large volumes of data
that are becoming available, there is the potential for making faster advances in many scientific
disciplines and improving the profitability and success of many enterprises.
Keywords: Big Data, Data Analytics, Challenges, Techniques, Oracle
1. Introduction
Big data is a new power that bring new data types and storage mechanisms. Big Data solutions are
ideal for analyzing raw structured data, also semi structured and unstructured data from a wide variety
of sources. Big Data platform gives the opportunity to extract insight from an immense volume,
variety, and velocity of data, in context, beyond what was previously possible. Three characteristics
define Big Data: volume, variety, and velocity. Big Data generally refers to data that exceeds the
typical storage, processing, and computing capacity of conventional databases and data analysis
techniques. As a resource, Big Data requires tools and methods that can be applied to analyze and
extract patterns from large-scale data. The rise of Big Data has been caused by increased data storage
capabilities, increased computational processing power, and availability of increased volumes of data,
which give organization more data than they have computing resources and technologies to process.
This paper begins by describing Big Data definition, characteristics and challenges. In section 3 we
discusses techniques for analyzing Big Data. Section 4 discuss the importance of Big Data. In section
5 we present a case study: Oracle Application in Social Insurance. Finally, the paper presents some
conclusions.
2. Big Data Definition, Characteristics and Challenges
Big Data is currently defined using three data characteristics: volume, variety and velocity. It means
that some point in time, when the volume, variety and velocity of the data are increased, the current
techniques and technologies may not be able to handle storage and processing of the data. At that
point the data is defined as Big Data. In the Big Data research, the term Big Data Analytics is defined
59
as the process of analyzing and understanding the characteristics of massive size datasets by extracting
useful geometric and statistical patterns. Big Data describes innovative techniques and technologies
to capture, store, distribute, manage and analyze megabyte or larger sized datasets with high-velocity
and diverse structures that conventional data management methods are incapable of handling.
Fig. 1. Four characteristics of Big Data: volume, variety, velocity and veracity.
2.1 Characteristics
The first characteristic of Big Data, which is “Volume”, refers to the quantity of data that is being
manipulated and analyzed in order to obtain the desired results. “Velocity” is all about the speed that
data travels from point A, which can be an end user interface or a server, to point B, which can have
the same characteristics as point A is described. “Variety” is the third characteristic of Big Data. It
represents the type of data that is stored, analyzed and used. The type of data stored and analyzed
varies and it can consist of location coordinates, video files, data sent from browsers, simulations etc.
“Veracity” is the fifth characteristic of Big Data and came from the idea that the possible consistency
of data is good enough for Big Data as is shown in figure 1. Current technologies software
technologies try to overcome the challenges that “V’s” raises.
2.2 Challenges
This section discusses the most salient of the challenges.
Privacy is the most sensitive issue, with conceptual, legal, and technological implications. The costs
of a data privacy breach can be enormous. Privacy is an overarching concern that has a wide range of
implications for anyone wishing to explore the use of Big Data for development.
Access and Sharing. Sharing the information proves to be one of the most valuable characteristics of
development. Every person and company has at their disposal large amount of information that can
use it to serve their purposes. Everything is available only if everyone shares it. Regarding persons,
there is a difference between what is personal and what can be made public. Working with new data
sources brings about a number of analytical challenges. The relevance and severity of those challenges
will vary depending on the type of analysis being conducted, and on the type of decisions that the data
might eventually inform. A new class of business software applications has emerged whereby
company data is managed and stored in data centers around the globe. While these solutions range
from ERP, CRM, Document Management, Data Warehouses and Business Intelligence to many
others, the common issue remains the safe keeping and management of confidential company data.
By storing data on cloud systems is more convenient for companies in terms of cost. Also, a cloud
system is not only characterized by storage space, but as well for the speed of processing requested
operations. The data security still remains a contentious issue in this case. These solutions often offer
companies tremendous flexibility and cost savings opportunities compared to more traditional on
premise solutions but it raises a new dimension related to data security and the overall management
60
of an enterprise’s Big Data paradigm. Since big data will lose its value to current decision-making
over time, and since it is voluminous and varied in content and structure, it is necessary to utilize new
tools, technologies, and methods to archive and delete big data, without sacrificing the effectiveness
of using your big data for current business needs.
3. Techniques for Analyzing Big Data
Big data analysis involves making “sense” out of large volumes of varied data that in its raw form
lacks a data model to define what each element means in the context of the others.
It is a daunting task in most industries and companies that deal with big data just to understand the
data that is available to be used, determining the best use of that data based on the companies’ industry,
strategy, and tactics. Also, these types of analyses need to be performed on an ongoing basis as the
data landscape changes at an ever-increasing rate, and as executives develop more and more of an
appetite for analytics based on all available information.
A common implementation that handles Big Data is MapReduce, presented in figure 2. MapReduce
consists of two things: mapping and reducing. This is a technique that programmers use when they
are confronted with large amount of data. By mapping a certain dataset is restructured into a different
set of values. Reducing is a process that takes several “mapped” outputs and forms a smaller set of
tuples.
Fig. 2. MapReduce
4. Importance of Big Data
When big data is effectively and efficiently captured, processed, and analyzed, companies are able to
gain a more complete understanding of their business, customers, products, competitors which can
lead to efficiency improvements, increased sales, lower costs, better customer service, and/or
improved products and services. The main importance of Big Data consists in the potential to improve
efficiency in the context of use a large volume of data, of different type. Big data is widely seen as
essential for the success of the design optimization of complex systems. Big Data can be used
effectively in the following areas:
61
In information technology in order to improve security and troubleshooting by analyzing the patterns
in the existing logs. In customer service by using information from call centers in order to get the
customer pattern and thus enhance customer satisfaction by customizing services. In improving
services and products through the use of social media content. By knowing the potential customers
preferences the company can modify its product in order to address a larger area of people. In the
detection of fraud in the online transactions for any industry. In risk assessment by analyzing
information from the transactions on the financial market.
5. Case Study: Oracle Application in Social Insurance.
Oracle comes with a complete system solution for companies. It starts from the basic ideas of Big
Data sources, which can be traditional data generated by ERP systems, sensor data and social data,
defined by the feedback that the company receives from customers. Oracle consists of Hardware and
Software from Oracle Corporation designed to integrate structured data and unstructured. This
integration includes the registration of data from different sources that generate massive volumes of
data. Also incorporated are different technologies to collect and analyze these data. A typical example
is the use of this application in Social Security. The goal is to collect as much data regarding the files
of pensioners. Registration of a folder starts from the registration of its request in ALSSH. It serves
as the initial source of information from the national register of citizens who called upon the data of
a person and his generalities, using its identification number. Another source of information is the age
of the work of a person who is taken from the archive of job years. Other sources of information are
all the documents submitted by the person who made the referral, where the documents are scanned.
Once registered applications and all documentation necessary for a person all along adopted through
an user confirmation and had to pass benefits specialist who accepts them and made relevant
calculations of the amount of pension in relation to claims made. Once the calculations are made
conform the beneficiary satisfies the conditions that made mass confirmation of the pension
calculation. Mean while after confirmation of the amount of pension becomes generation (transfer
payment). The latter enables online over payment of social insurance in payment centers. Data
collection consists of data which are provided by the National Register of Citizens, which is connected
online to the civil state, and the data which are mainly provided by the beneficiaries of retirement.
Figure 3 shows the view from user accounts that make registration of records in this application.
Fig. 3. Online user registration
62
Conclusions
Big Data concept and the technologies associated in order to understand better the multiple benefices
of this new concept ant technology. Big data has to do with volume as well as the speed of data
storage. Since we are dealing with a transaction that efficiency leads to improvements in our services
offering significant advantages. Big data allows organizations to create highly specific segmentations
and to tailor products and services precisely to meet those needs. Big data enables companies to
create new products and services, enhance existing ones, and invent entirely new business models.
References
1. P. C. Zikopoulos, C. Eaton, D. deRoos, T. Deutsch, and G. Lapis, Understanding big data –
Analytics for enterprise class Hadoop and streaming data, McGraw-Hill, 2012.
2. P. Zikipoulos, T. Deutsch, D. Deroos, Harness the Power of Big Data, 2012,
3. Oracle Information Architecture: An Architect’s Guide to Big Data, An Oracle White Paper in
Enterprise Architecture August 2012
4. McKinsey Global Institute - Big data: The next frontier for innovation, competition, and
productivity – June 2011
5. T. Chai, Y. Jin, and S. Bernhard, “Evolutionary complex engineering optimization: Opportunities
and challenges,” IEEE Computational Intelligence Magazine, vol. 8, no. 3, pp. 12–15, 2013.
6. Y. Noguchi. The Search for Analysts to Make Sense of Big Data. National Public Radio
(http://www.npr.org/2011/11/30/142893065/the-search-foranalysts-to-make-sense-of-big-data),
Nov 2011.
7. J. Manyika, M. Chui, B. Brown, J. Bughin, R. Dobbs, C. Roxburgh, and A. H. Byers. Big data: The
next frontier for innovation, competition, and productivity. McKinsey Global Institute, May 2011.
63
Assessing Clustering in a Social University Network Course
Orgeta Gjermëni
Department of Mathematics, University ʺIsmail Qemaliʺ
Vlore, Albania
o.gjermeni@gmail.com
Abstract. A collection of data is gathered from surveys held in a Spring Course of the Economic
Faculty in the University “Ismail Qemali” of Vlora, Albania. The data set for each student contains
the names of the other students through which he/she have a “social relationship”. This relationship
includes frequent communications, discussions on exercise solutions, and sitting usually close to each
other in the class. We have constructed. At the end of the course, a final network based on this type
of relationship. We are particularly interested on the clustering coefficient of this network and
assessing it’s “significance”, in the sense of being somehow unusual or unexpected. Simulated
random graph models, using R platform, are used to test the “significance” of the observed clustering
coefficient.
Keywords: Social networks, Clustering coefficient, Random graph models
1. Introduction
Relationships and collective behavior in many fields are popularly studied as part of network analysis
[11 – 13]. Real world networks can be seen as complex systems, which are unstable, multi –
parameter, nonlinear and time varying. Information flows about jobs, new products, technologies or
political opinions, spread of epidemic diseases are some examples of social networks [1, 2] effects in
our lives. The concept of clustering in social networks was formally introduced by Watts and Strogatz
[3], but also in sociology [2]. Clustering coefficient, known also as transitivity, as a global measure,
summarizes the relative frequency with which connected triples close to form triangles. It gives
answer to the question: – How well vertices in a network tend to be clustered together? Also,
clustering coefficient may be seen as the probability that two nearest neighbors of a vertex are also of
one another. The motivation for this interest in clustering coefficient is because in real life, our beliefs,
decisions and behaviors are constantly influenced by people with whom we regularly or sporadically
interact. Clustering coefficient plays an important role as a statistic parameter on analyzing those
obs
networks. For every observed network, G
, it is possible to find its clustering coefficient, cl (G obs )
, but our purpose in this study is on assessing whether it is ‘significant’, in the sense of being somehow
unusual or unexpected. In this paper we analyze a social network, conceived as a fixed set of vertices
and edges from the perspective of clustering coefficient of the network. Each vertex represents a
university student, and each edge represents a ‘social relationship’. The relation is defined by
frequently communications, discussions on exercise solutions, or sitting usually near each other in the
classroom.
The reminder of this paper is organized as follows: In Section 2, we describe the method that is
applied. Next, in Section 3, we see the results, and in Section 4, we discuss about the results. Finally,
we present some concluding remarks and future work in Section 5.
2. Methods
Data was obtained from surveys which were held periodically during a Spring 2015 – semester course
(15 weeks) of the second year students at the Economic Faculty in the University “Ismail Qemali” of
Vlora. For each semester course, students have the possibility to choose the lecturer they want
64
between some alternatives. A ‘mixing process’ happens at the beginning of every course within the
various groups of students.
During the surveys, each of the students showed the names of other students, for which he/she had a
‘social relationship’ with. This social relationship includes frequent communications, discussions on
exercise solutions, and sitting usually close to each other in the class. In this way is defined the
socialization that happens within the course.
A social relationship is assumed to be ‘forever’ from the moment it starts between two students, till
the end of the course. Based on the conducted surveys, a final social network graph is conceived as a
fixed set of vertices and edges G
network graph is simplified.
obs
 (V
obs
,E
obs
)
. The orientation of the edges is neglected and the
Since for every observed network graph, it is possible to find its clustering coefficient, cl (G obs ) , in
our study we are interested in assessing whether this value is ‘significant’, in the sense of being
somehow unusual or unexpected. Hypotheses we arise here is:
H 0 : G obs can be viewed as a uniform sample under either random graph model.
Network graph models are used for a variety of purposes, also for testing the significance of a pre –
defined characteristic in a given network graph. Significance is defined in comparison to an
appropriate form of reference. Random graph models [4] are used in setting up such comparisons.
Therefore, for our purpose we will define two collections:
obs
1
– Collection of random graphs of the same order and size as G , classical random graphs;
obs
2
G
– Collection of random graphs with the same degree distribution as
graphs.
, generalized random


Let P() be a uniform probability distribution over 1 , and 2 . The value of cl (G obs ) is compared
to the collections of values cl (G ) : G   and cl (G ) : G  2 . If cl (G obs ) is judged to be extreme
1




with respect to this two collections, than that is taken as evidence that G
obs
is unusual in having this
value [5, 6]. This evidence will contradict the null hypothesis H 0 . Simulations are conducted through
“vl” method [7] as a generator and Monte Carlo algorithms are used for randomizations. This is
implemented in R [9], package igraph [8].
65
3. Results
Fig. 2. This shows a visualization of the Social University Network Course, Gobs using layout
generated by the Kamada – Kawai algorithm [10]. The size of each vertex is specified by its degree.
In the simplified observed network graph were found a total of 89 vertices and 494 edges. It resulted
connected and the clustering coefficient was cl (G obs )  0.375 . A visualization of the Social
University Network Course (the observed network graph) is given in Fig.1.
After simulating two uniform samplings of 10.000 random graphs for each of the collections  and
1
2 , clustering coefficient was calculated for each graph sampled. Histograms in Fig. 2 show a
summary of the resulting clustering distributions.
Fig. 2. Histogram distributions of clustering coefficients cl (G ) detected for random graphs, G
generated uniformly from 1 (same size and order as the observed graph) and 2 (same degree
distribution as the observed graph).
66
4. Discussion
Clustering coefficient is considered significantly high if cl (G obs )  cl (G ) , where G is a random
graph from 1 or 2 . Based on the results, we can say that there is a strong evidence to reject the null
hypothesis H . Clustering coefficients values in both simulated collections are relatively small
0
compared to the observed clustering network graph value.
The so called ‘karate club network’ of Zachary [14], has a somewhat similar nature with our network.
The clustering coefficient of the ‘karate club network’ was 0.2556818 [15]. Compared to this value,
our observed network clustering value is greater.
Conclusions
As a conclusion, at the end of this paper we can say that the observed network graph showed
significant greater clustering coefficient compared to that of random graph models with a comparable
magnitude (i.e., with respect to order and size) or comparable connectivity (i.e., with respect to degree
distribution). We can’t assume that our network graph is a uniform sample under either random graph
model. Further investigations should be done on analyzing average path length and its ‘significance’
with the purpose detecting ‘small world’ [3] behavior in this network graph.
Acknowledgments go to the students of Spring 2015 – semester course of the second year –
Economic Faculty at the University “Ismail Qemali” of Vlora for their collaboration on sharing
personal data.
References
1. Scott, J.: Social Network Analysis: A Handbook, Sage Publications, London, 2 nd ed. (2000)
2. Wasserman, S., Faust, K.: Social Network Analysis, Cambridge University Press, Cambridge
(1994)
3. Watts, D. J., Strogatz, S. H.: Collective Dynamics of ‘Small – World’ Networks, Nature, Vol. 393.
pp. 440 – 442, Macmillan Publishers Ltd (1998)
4. Erdös, P., Rényi, A.: On Random Graphs, Publicationes Mathematicae, Vol. 6, pp. 290 – 297 (1959)
5. Kolaczyk, E. D., Csárdi, G.: Statistical Analysis of Network Data with R, Use R! 65, DOI:
10.1007/978 – 1 – 4939 – 0983 – 4_5, Springer Science + Business Media, New York (2014)
6. Kolaczyk, E. D.: Statistical Analysis of Network Data with R, Springer Series in Statistics, DOI
10.1007/978 – 0 – 387 – 88146 – 1_6, Springer Science + Business Media LLC (2009)
7. Viger, F., Latapy, M.: Fast Generation of Random Connected Graphs with Prescribed Degrees,
<hal - 00004310>, (2005)
8. Csárdi, G., & Nepusz, T.: The igraph Software Package for Complex Network Research,
InterJournal, Complex Systems 1695. http://igraph.org/, (2006)
9. R Core Team R: A Language and Environment for Statistical Computing, R Foundation for
Statistical Computing, Vienna, Austria. http://www.R-project.org/, (2015)
10. Kamada, T., Kawai, S.: An Algorithm for Drawing General Undirected Graphs, Information
Processing Letters, Vol 31(1), pp. 7–15 (1989)
11. Albert, R., Barabási, A.-L.: Statistical Mechanics of Complex Networks, Reviews of Modern
Physics, Vol 74 (1), pp. 47 – 97 (2002)
12. Dorogovtsev, S. N., Mendes, J. F. F.: Evolution of Networks, arXiv: cond-mat/0106144v2[condmat.stat-mech](2001)
13. Boccaletti, S., Latora, V., Moreno, V., Chavez, M., Hwang, D. – U.: Complex Networks: Sturcture
and Dynamics, DOI:10.1016/j.physrep.2005.10.009 (2006)
14. Zachary, W.: An Information Flow Model for Conflict and Fission in small Goups, Journal of
Anthropological Research, Vol 33(4), pp. 452 – 473 (1977)
67
15. Kolaczyk, E. D., Csárdi, G.: Statistical Analysis of Network Data with R, Use R! 65, DOI:
10.1007/978 – 1 – 4939 – 0983 – 4_4, Springer Science + Business Media, New York (2014)
Perfect Metamaterial absorber based energy harvesting
application in ISM Band
Furkan Dincer1, Mehmet Bakir2, Muharrem Karaaslan2, Kemal Delihacioglu1, Cumali
Sabah3
1
Department of Electrical and Electronics Engineering, Kilis 7 Aralik University, Kilis,
Turkey
2
Department of Electrical and Electronics Engineering, MTMs and Photonics Research
Group Iskenderun Technical University, Hatay, Turkey
3
Department of Electrical and Electronics Engineering, METU - Northern Cyprus
Campus TRNC / Mersin 10, Turkey
furkandincer@kilis.edu.tr
Abstract - An electromagnetic (EM) energy harvesting application based on metamaterial absorber
(MA) is introduced and discussed in this paper. This application is operating at ISM band (2.40 GHz)
which is especially chosen due to its wide usage area. Square ring resonator (SRR) which has two
gaps and resistors across the gaps on it is used. Chips resistors are used to deliver power to any active
component. Transmission and reflection characteristics of metamaterial absorber (MA) for energy
harvesting application are investigated. 83.6% efficient for energy harvesting application is realized
in this study.
Keywords: Metamaterials, Absorber, Energy harvesting
1. Introduction
Metamaterials (MTMs), due to their exotic application areas attract attention in recent years. These
applications can be categorized as follows. Cloaking [1], antennas [2], sensors [3-4], absorbers [5, 6],
etc. [7-8].Nowadays, new applications by using split ring resonators (SRRs) in MTMs came out [910]. When electromagnetic waves excite SRRs, generated electromagnetic energy is absorbed by
resonators. This energy can be used as a voltage source, In order to harvest the energy, a gap is
positioned on SRR coherent with the electric field direction and lumped element is placed on a gap to
harvest this energy. Lumped element can optionally be a varicap or resistor. This mechanism is a basis
for energy harvesting applications. If a lumped element is placed across the gap such a varicap, tunable
harvesting applications can be possible. In this study,the lumped elements are choosen as resistors
which are the recipient of the harvested energy. Resistance value is determined according to the
numerical results. EM Energy harvesting concept is new technology for MTM science, but it has big
importance due to limited energy resources and big field of applications.
Power harvesting devices are generally used to convert one type of energy as acoustic or optical to
another type of energy which is generally DC [11-12]. Though EM MTMs provide flexible design
options and they can be adapted to various energy harvesting studies, there is few electromagnetic
energy harvesting studies with MTMs in current literature [13-15]. At first, 611mV voltage is
harvested at 5.8GHz by using electromagnetic MTMs[13]. Energy conversion by using 3-D
metamaterial arrays are presented in [14], at [15], 36.8% power efficiency is experimentally obtained
at 900MHz.
This study has different sides and advantages from the others as explained below. First advantage of
this study is having simple geometry. Simple geometry means simple production, anyone who wants
to produce this MTM energy harvester can produce easily. Other advantage is that; this study based
on metamaterial absorber (MA) together with tunability option. It is possible to get more than 83.6%
68
efficient MTM energy harvester with this option. Other advantage of this study is the realization for
ISM band. Lots of appliances are operated at this band so it can be possible to harvest energy without
making extra transmitters. The organization of this study is as follows. In Section 2, the structure
geometry based on SRR is proposed, and the numerical method is presented. In Section 3, numerical
results are presented. Finally, summary and conclusions are provided and discussed in Section 4.
2. Design and Numerical
In this study, a SRR as an example of resonant MTM particle is used as a MA based energy harvester.
Square ring resonators (SRR) used in this design have two 7.2mm gaps. Resistors are placed as
lumped network elements across the gaps on SRR. Resistance value is chosen as 2000Ω which will
be explained later. FR4 material is chosen as a substrate to support resonator and reduce resonance
frequency as well as frequency selective surfaces. FR4 substrate has a thickness of 1.6mm with loss
tangent and relative permittivity of FR4 0.02 and 4.2, correspondingly. Copper is used as metallic
element with a thickness of 0.035mm and it has a conductivity of 5.8𝑥108S/m. An air gap which has
a thickness of 4.2mm is placed behind the substrate. Back of dielectric-air gap, copper plate is placed
as an exact reflector to block signal transition. Copper plate and air gap dimensions are hold same
with FR4 substrate dimensions. Design details can be seen in Fig. 1. Numerical results are achieved
by commercially available full-wave electromagnetic solver which uses finite integration technique.
Fig. 1. Dimensions of the MA for energy harvesting study.
The frequency response of absorber can be calculated by A(ω) = 1 − R(ω) − T(ω), where
A(ω), 𝑅(ω) and 𝑇(ω)define the absorption, reflectance, and transmittance, respectively. When
reflection R(ω) = |S11 |2 and transmission T(ω) = |S21 |2 falls to minimum values, absorption rises
to maximum at the resonance frequency. There will be no transmission to be examined since a metal
plate is assigned behind air type dielectric.
The perfect absorption happens when the reflection is close to zero. When perfect absorption is
observed power harvested from resistors rises which is crucially important for energy harvesting
studies. It is possible to reduce reflection to near-zero by choosing the relative effective permittivity
𝜀(𝜔)and permeability 𝜇(𝜔)values closer to provide impedance matching with air. It is possible to
absorb both the incident electric and magnetic field by properly tuning 𝜀(𝜔) and 𝜇(𝜔)of the effective
medium of absorber. There are some studies in literature to explain absorption mechanism of the
structures [15-18].
3. Results and Discussion
In order to show the performance of the proposed MA based energy harvester, numerical study is
realized and assessed. Numerical values of reflection and absorption are given in Fig. 2. It is seen
from Fig.2 that the absorption and reflection values are 99.99% and0.02%, respectively when2000Ω
resistors are placed in gaps. According to these values suggested MA structure shows perfect
absorption features at 2.40GHz.
69
Fig. 2. Absorption and reflection value.
One of the most important feature of power harvesters is RF to DC power conversion. Power
conversion can be calculated as [13];
𝜌=
𝑃𝐷𝐶
𝑃𝑅𝐹
(1)
𝑃𝑅𝐹 isdefined as 1𝑊 by simulation program manufacturer. 𝑃𝐷𝐶 values are extracted from voltage and
current values of resistors used in this study.Resistors are placed across the gaps of SRR to obtain and
easily convert voltage value using conventional methods.The resistance value has a critical
importance and a great effect on the total energy conversion efficiency of the structure. In order to
show the harvested power performance for MA based harvesting application, the effects of resistance
value on power harvesting efficiency is investigated for comparison. This comparison enables a
deeper understanding for operating mechanism of the suggested structure. The efficiency is examined
for ten different resistance values between 500 Ω𝑎𝑛𝑑 5.000Ω Numerical results are given in Fig. 3.
Fig. 3. RF to DC efficiency of SRR according to load resistance
As shown in Fig. 3, changing the resistance values between 500 Ω𝑡𝑜 2.000Ω results in significant
change in power harvesting efficiency between 58% 𝑡𝑜 83.6%. When 2 𝑘Ω resistance is used,
0.418 𝑊power is obtained from a single resistor which means 0.836 W power collection by using
two resistors. For this reason overall efficiency of MA energy harvester becomes around 83.6 %
𝑃
according to the equation of 𝐷𝐶 . This is the highest energy harvesting value in the literature. Beside
𝑃𝑅𝐹
this, the efficiency of the proposed structure linearly changes depending on the load resistance 2 𝑘Ω −
5𝑘Ω. Hence the system can also be adopted to many sensor applications of which resistance changes
linearly with respect to the intended measurement parameters.
Table 1. Energy Harversting Efficiency Comparison Table (Single Array)
Reference study
Efficiency (%)
Ramahi [13]
76%
Almoneef [14]
68%
Hawkes [15]
65%
Proposed Study
83.6%
70
Metal plate which is behind the air gap is acting important role in MA based harvesting studies. EM
energy is confined between metal plate and resonator which increases the power across the resistors.
A numerical application is realized and investigated for the proposed MA based energy harvesters.
Two different configurations (with-metal plate and without-metal plate) are created and investigated
for resistance value of2 𝑘Ω. First investigation for this numerical setup is related with absorption rates.
As it seen from Fig.4, when metal plate is present behind the air gap, absorption value becomes
99.99% at 2.40 𝐺𝐻𝑧. In contrast to the first model, when metal plate is not used, absorption value
drops dramatically to 44 %at2.46 𝐺𝐻𝑧.
Fig. 4. Absorption Rate of MA base energy harvesting studies according to existence of metal plate.
Second investigation is related with power and resonance frequency according to the existence of
metal plate. For the first case as it seen from Fig. 5 that power value which is harvested from a single
resistor is 0.418 W at 2.40 GHz. However it is 0.11W at 2.46 GHz for the without-metal plate model.
Metal plate affects the resonance frequency and absorption level directly because it creates EM
effects. It can be seen that the position of the resonance is shifted from 2.40 𝐺𝐻𝑧 to 2.46 GHz when
metal plate is removed. This means that each case has a different characteristic, capacitive effect and
power level. Also, the metal plate keeps the EM transmission value near zero. It is concluded from
this numerical analysis that adding metal plate in MTM energy harvesting studies increases the power
harvesting efficiency in high amounts. As shown in the table 1, when this work is compared with
other studies in current literature, efficiency reaches maximum value in this study. Hence, MTM based
harvester is a novel and challenging study in literature with its better results and primness.
Fig. 5. Power harvested across the lumped element according to existence of metal plate.
Physical mechanism and operation principle can be understood easily according to electric field
distributions of front and side metallic parts of the absorber. Electric field distributions are given at
the resonance frequency of 2.40 GHz. It can be seen from Fig.6. that the electric field is concentrated
on SRR and it is confined due to metal plate. By adding a resistor to the resonator, the structure can
be modeled as a voltage source. In order to realize this, structure is modeled as perfect absorber which
provides maximum EM energy harvesting at the resonance frequency.
71
Fig. 6. Electric field distribution at resonance frequency.
Conclusion
In this study, an energy harvesting application operating at ISM band is proposed and discussed.
Electromagnetic characteristics of the proposed design and usage areas in EM energy harvesting
applications are discussed. Efficient pPower harvesting with 83.6% efficient at 2.40 GHz is
demonstrated numerically. Numerical results prove that this study can be used in energy harvesting
application efficiently. As a result, proposed structure can be used as MA based energy harvester with
high efficiency in ISM band.
References
1.
Pendry, J. B., Schuriand, D., Smith D. R.: Controlling electromagnetic fields Science, Vol. 312.
(2006) 1780-1782
2. Li, D., Szabo, Z., Qing, X., Li, E. P., Chen, Z. N.: A high gain antenna with an optimized
metamaterial inspired superstrate IEEE T. Antenn. Propag, Vol. 60. (2012) 6018-6023
3. Sabah, C., Dincer, F., Karaaslan, Unal, M. E., Akgol, O., Demirel, E.: Perfect metamaterial
absorber with polarization and incident angle independencies based on ring and cross-wire
resonators for shielding and a sensor application Opt. Commun, Vol. 322 (2014) 137-142
4. Shelby, R.A., Smith, Schultz, D.R.:Experimental verification of a negative index of refraction,
Vol. 292. Science (2001) 77-79
5. Chen, H.T.: Interference theory of metamaterial perfect absorbers, Opt. Express, Vol. 20. (2012)
7165-7172
6. Lee, J., Lim, S.: Bandwidth-enhanced and polarization-insensitive metamaterial absorber
usingdouble resonance Electron. Lett, Vol.47 (2012) 8-9.
7. Esen, M. Ilhan, I., Karaaslan, M., Unal, E., Dincer, F., Sabah, C.: Electromagnetic absorbance
properties of a textile material coated using filtered arc-physical vapor deposition method, J. Ind.
Text, (2014) 1528083714534710
8. Dincer, F., Karaaslan, M. Unal, E. Akgol, O., Demirel, E., Sabah, C.:Polarization and angle
independent perfect metamaterial absorber based on discontinues cross-wire-strips,”J.
Electromagn. Waves Appl, Vol. 28. (20149 741-751.
9. Lee, H.J., Yook, J.G.:Biosensingusingsplit-ring resonator at microwaveregime, Appl.
Phys.Lett.,Vol.92. (2008) 254103:1–254103:3
10. Ou, J.Y., Plum, E., Jiang, L., Zheludev, N.I.: Reconfigurablephotonic metamaterials, Nano Lett,
Vol.11. (2011) 2142–2144,
11. Lallart, M., Guyomar, D. Richard C., Petit, L.: Nonlinear optimization of acoustic energy
harvesting using piezoelectric devices, J. Acoust. Soc. Am., Vol.128. (2010) 2739–2748
12. Zhu, N., Ziolkowski, R. W., Xin, H.: A metamaterial-inspired, electrically small rectenna for
high-efficiency low power harvesting and scavenging at the GPS L1 frequency, Appl. Phys.
Lett., (2011) Vol. 99. 114101
72
13. Ramahi,O. M., Almoneef, T. S., AlShareef, M., Boybay, M. S.:Metamaterial particles for
electromagnetic energy harvesting, Appl. Phys. Lett. (2012) Vol. 101. 173903
14. Almoneef, T. S., Ramahi, O. M.: A 3-dimensional stacked metamaterial arrays for
electromagnetic energy harvesting, Prog. Electromagn.Res., (2014) Vol. 146. 109-115
15. Hawkes, A. M., Katko, A. R., . Cummer, S. A.: A microwave metamaterial with integrated power
harvesting functionality, Appl. Phys. Lett., Vol. 103. (2013) 163901
16. Dincer,F., Akgol,O., Karaaslan, M., Unal, E., Sabah, C.: Polarization angle independent perfect
metamaterial absorbers for solar cell applications in the microwave, infrared, and visible regime,
Prog. Electromagn.Res, Vol. 144. (2014) 93-101,
17. Sabah, C., Dincer, F., Karaaslan, M., Unal, E., Akgol, O.: Polarization-Insensitive FSS based
Perfect Metamaterial Absorbers in GHz and THz Frequencies, Radio Science, Vol.49. (2014)
306-314
18. Dincer, F., Karaaslan, M., Unal, E., Akgol, O., Sabah, C.: Design of Polarization- and Incident
Angle-Independent Perfect Metamaterial Absorber with Interference Theory, Journal of
Electronic Materials, Vol. 43. (2014) 3949-3953
73
Web Application for Students and Lecturer of Electrical
Engineering Faculty
Genci Sharko1, Anni Dasho Sharko2, Kadri Manaj1
1
2
Polytechnic University of Tirana, Electrical Engineering Faculty
Europian University of Tirana, Faculty of Economy and Information Technology
{gsharko1, anidasho2}@gmail.com
Abstract. Electrical Engineering Faculty, nowadays needs to develop by their students the “FIE
Student MIS Application”. Development and usage of Web applications are popular due to the
ubiquity of the browser as a client, commented mostly as a thin client. The ability to update and
maintain web applications without distributing and installing software on potentially thousands of
client computers is a key reason for their popularity. This paper presents the “FIE Student MIS
Application” web platform designed for Electrical Engineering Faculty students and their lecturer to
support on time correctly without delays and informing online all students and their lecturer receiving
online the student’s status as per the FIE exam schema: attendance/activation at lecturer/seminar
presence %, exam %, project/essay %, final grade of the exam; schedule of lecturers/seminars by
day/week/month per class, other information coming from secretary part to all students in general or
to one student. The process of implementation of the “FIE Student MIS Application” is based on
project management of web application as per: definition, planning, management and evaluation.
Keywords: web applications, project management, PhP and MySQL, Student MIS Application.
1. Introduction
Development, implementation, and usage of web applications are having a great popularity due to the
ability of updating and maintaining web applications without distributing and installing software on
all Electrical Engineering Faculty student’s computers of Polytechnic University of Tirana for the
whole set of applications developed for FIE students’ needs and FIE Managerial and Academic staff
needs. Web-based applications are: easier to be developed from the development teams, more useful
for your users, userfriendly applications, easier to install and maintain and keep secure the whole
application already developed, easier to grow as the needs from university come and grow time by
time. The Project Management process of Developing Web Applications is based on: Analyses,
Development, Testing, Implementation and Usage of Web Applications which are having a great
popularity even due to the ability of updating and maintaining web applications without distributing
and installing software on potentially all FIE student’s computers of Polytechnic University of Tirana.
[1] “FIE Students-MIS” web application is the first initiative of web application designed for
Polytechnic University of Tirana, Electrical Engineering Faculty, from Electrical Engineering Student
where the outcome of students’ and lecturers’ intranet platform is seen from FIE Management,
lecturers and students as an important tool facilitating students and their lectures during the academic
year with: Course’s syllabuses, lecturer’s presentations in power point format, downloaded e-books,
mandatory and recommended literature for the courses, video tutorials for different subjects.
The on-line evaluation platform “FIE-On-line lecturers/courses evaluation” web application is the
first initiative of web application designed for Electrical Engineering Faculty by electrical engineering
students where the outcome of students’/courses evaluation of teaching performance is seen from FIE
Management as an important tool to measure the effectiveness of teaching quality of all
lecturers/courses for the whole academic staff bachelor and master levels of Electrical Engineering
Faculty.
74
2. Web Application Analysis and Definitions
A Web-based application refers to an application that uses the World Wide Web as a distributed
informal retrieval system. The writing process of web applications is often simplified by open source
software (Drupal, Ruby or Symphony) so often called web application frameworks. These web
application frameworks facilitate the rapid application development process allowing IT development
team to focus on the parts of the application which are unique to their goals without having to resolve
common development issues such as user management/reporting part. (Lane,Williams, 2004)[3],
(Greenspan,Wall,Bulger, 2004)[4].
An intranet is a network that provides services within an organization/university environment that are
similar to those provided by the Internet platform, but those services are not mostly connected to the
Internet platform. Intranets are typically protected by “firewalls” based on the security policy of the
organizations/university environment, where an intranet can be considered as a private internet. Webbased applications can be developed flexibly on both the Internet and Intranets platform of an
organization/university environment. A web-based applications typically use web browsers Mozilla
Firefox or/and Microsoft Internet Explorer, (mostly we use at our web applications platform Mozilla
Firefox) as a platform and HTTP as an additional Internet protocol.
The detailed analyses consisted on several meetings and interviews done with the Faculty Dean and
Deputy Dean, Head of Departments the requirements specification document was created for
developing web applications as:
“FIE Students-MIS”
“FIE-On-line lecturers/courses evaluation”
Review of existing FIE official documentations helped us on understanding all elements of actual online and traditional hardcopy evaluation process and Intranet documentation platform. The FIE
student’s main database platform already existing in excel format, was a good reference when working
for the final requirements specification document for “FIE-Online Lecture/Courses Evaluation”, and
“FIE Students-MIS”, Web applications.
Project Methodology used:
First phase had the following sub-phases:

Analysis and Definitions

Implementation (Active development)

Backup Policy (daily, weekly during the evaluation period per each semester).
Second phase had the following sub-phases:

Final preparation (Migration, Integration tests, Documentation, user training)

Go Live in 2 phases

First Phase testing with Bachelor Students of Electrical Engineering Faculty, during the
end of the Electrical Measurement Subject for both applications.

Second phase Go-Live with full versions of both web applications.
In general, all FIE students accessing the application form Intranet or Internet environment found the
importance of such application and how useful was for all students both web platforms: FIE-Online
Evaluation and FIE-Students MIS. The success or the failure of a software system depends mostly
on its utilization. (Zhiming,Woodcock,Huibiao, 2013)[5], (Sommerville, 2012)[6].
3. Software Design, Data and Software Model
3.1 “FIE-Students MIS” Application
Building up a Management Information System for university environment for sure it needs a unique
combination of features. Based on that thinking about building an Students and Lecturers MIS for
Electrical Engineering Faclty, Polutechnic University of Tirana Environment, several requirements
will come to the fore: to decide on what platform works best for us, we need to analyze these
requirements, then evaluate systems based on on-site cost development & maintenance.
As we all know the concept of Intranet, even for a university environment we have the same
properties: private network, admission across firewall, requires authorization, useful for delivering &
sharing information on selective basis. The FIE developer’s team went through a deep analyses phase,
75
taking into consideration the business requirements of FIE Management and Academic Staff and
concluded as hereby:
 FIE Students-MIS should be developed over an open source platform, Joomla or PhP MySQL.
 This system will be able of creating possibilities of adding different items, during the extension
of the Project second phase.
 Eligibility for managing FIE student’s access (as user’s access on the system).
 The system should give access materials by students on the basis of group-level subjects (ie
Electrical Engineering Faculty Master Students can access and download materials like lectures
in ppt, e-books, other video materials and articles related to the subject. Students will not access
other Departments, of the Electrical Faculty or other master or bachelor profiles even inside the
Electrical Engineering Faculty if they do not belongs to.
 Lectures of the subjects should have the right to upload materials (lectures in ppt, e-books
(mandatory & recommended literature), only in their subject’s folder/semester/academic year and
can update each academic year the infrastructure of materials entered supporting their students.
 The FIE Students and Lecturers MIS Administrator has the rights to enter users as per students
increase on each master or bachelor level.
 Students can manage their account with details on their presence on the lecturing for the whole
simester, project evaluation, exam evaluation, and final evaluation in points and in degree as well.
 They will find there the whole week lecture, seminars and laboratories to be present during hours,
days and weeks till the end of the simester and academic year.
Building up the FIE Students & Lecturer MIS platform: we decided to use the FIE Internet template
and to go further on customization and developing to achieve FIE Intranet objectives. For creating the
subjects folder, in the menu content management, in the administration panel: it can create categories,
add articles etc. Creating Users will only be made by a panel of administration in order to eliminate
the creation of accounts by unauthorized persons. Originally must create user groups and then levels
of access.
Figure 1. FIE Students and Lecturers MIS.
The second level of access will be the Electrical Engineering Faculty, as group of users as the first
main template to be used after for the other faculties of the whole Polytechnic University of Tirana as
well on the second phase of the project implementation. The FIE Intranet administrator will create an
account for one student of Master Degree „Elektroteknike/Sistemet Elektrike te Fuqise/Automatizimi
i Industrise“ - Electrical Engineering Faculty, he will choose the Electrical Engineering Faculty as
group of users-first level of access.
76
Figure 2. The main page of Students accessing as per their credential the FIE Students MIS
platforme.
Figure 3. FIE Students Subjects Evaluation Process.
Figure 4. FIE Students Lecturer-Seminars-Laboratiry to be present per hours/days/weeks.
77
3.2 “FIE-Online Lecture/Courses Evaluation” Application
For FIE On-Line Evaluation Application we have chosen the WAMP platform as a dynamic webbased application of type client-server developed and implemented in a Windows-Apache-MySQLPHP platform. The application is installed once in the server and, all students, lecturer, and the top
management staff of FIE can access and use the application.
Realizing such a web application for on-line evaluation process for the lecturing process and courses,
the FIE IT team designed this web application based on the platform Apache/MySql/PHP with
following versions of the components: Apache 2.2.14, MySql 5.1.44, PHP 5, PhpMyAdmin 3.2.4.
Properties of FIE-Online Lecture/Courses Evaluation application are:
 The database of this application will be in MySql, in the next step it will be on SQL.
 The on-line evaluation application will be available over the FIE Intranet/Internet platforms.
 The web application is not depandable from the operating system of the server.
 The application can be accessed from different browsers, but preferable will be Mozilla as per
FIE standard web page browser.
 The users have different rights and not overlapping with each other.
 The basic reports is based on the functional requirements elements.
 Automatic email to be sent to all FIE lecturer/ management staff after the whole evaluation
process is finalized.
During the designing of the “FIE-Online Lecture/Courses Evaluation” Database, we have done the
proper analyses based on the main FIE Students Database and a well-designed database took time and
efforts to conceive, build and refine, to achieve at the end the proper database structure for the “FIEOnline Lecture/Courses Evaluation” Application.
Fig. 5. Login form for the FIE Students.
Fig. 6. Password generation based on FIE student’s ID and email address.
78
Fig. 7. Lecturer and Course Evaluation form to be fill in from all students online.
An effective data model completely and accurately represents the data requirements of the end users.
The model used for “FIE-Online Lecturers/Courses Evaluation” Application eliminates redundant
data, it is independent of any hardware and software constraints, and can be adapted to changing
requirements with a minimum of effort.
After the analyses phase based on the collaboration of functional and IT it was agreed for the further
steps on the software design part as usually considered as the most crucial one or the success of a FIE
On-Line evaluation Application development part. It consists in developing a database and software
model which will interact with the database for storing, modifying and retrieving proper data. [2]
The first step in this phase was modeling the real world system in a well-structured database. The next
step is the implementation of the software for interacting with the database and most importantly
offering a user friendly interface to do so.
The communication between the database and the software includes: storing data/information into the
database, modifying data/information already stored in the database, retrieving and consulting
data/information. Each user of the application should fill identification requirements in order to login
in its personalized interface and use the application and its features. (Runeson, Host, Rainer, Regnell,
2012)[7], (Bell, 2005)[8].
79
Conclusion
As we all know web-based applications are well-known for easier to be developed from the IT
development teams created mos of the times with faculty bachelor and master students at FIE, are
more useful for all type of users: students, lecturer or administrative university staff. Easier to be
installed and maintained and with high level of security regarding the whole application already
developed by the IT development team, easier to be grown as the needs from university come and
grow time by time. With the initiative of the Electrical Engineering Faculty Management and
Academic Staff it was established a dedicated team of a bachelor student mentored from the Deputy
Dean for the development of two web development applications necessity supporting and facilitating
students and lecturers during the academic year: a.“FIE-On-line lecturers/course evaluation” through
the desktop application platform b.“FIE Students & Lecturer- MIS” supporting FIE students and their
lecturers during the academic year.
Encouraging FIE students’ community acceptance and adoption of Web applications calls for
initiatives to make such applications more broadly useful to users in their daily activities. To this end,
we claim that a path - based incremental development approach, in which users are involved in
evaluating each increment, is a good approach for appropriate technology Web applications. The
development of this application is done based on the following platforms of: Apache/MySQl/PHP.
References
1. Anderson, J. & Brown, G. & Spaeth, S. (2006). Online student evaluations and response rates
reconsidered, Innovate: Journal of Online Education, 2(6).
2. Thorpe, S. W. (2002 June). Online student evaluation of instruction: An investigation of nonresponse bias. Paper presented at the 42nd Annual Forum of the Association for Institutional
Research, Toronto, Ontario, Canada.
3. Lane, D. and Williams, H. E., (2004). Web Database Applications with PHP and MySQL. O'Reilly.
4. Greenspan, J., Wall, D. and Bulger, B. (2004). MySQL/PHP Database Applications. Wiley.
5. Zhiming, L. Woodcock, J. & Huibiao, Zh. 2013. Unifying Theories of Programming and Formal
Engineering Methods. International Training School on Software Engineering Held at ICTAC.
6. Sommerville, I. 2012. Software engineering 9th ed. Addison-Wesley, Pearson Education.
7. Runeson,P. & Host, M. & Rainer, A. & Regnell, B. (2012). Case Study Research in Software
Engineering (Guidelines and Examples). John Wiley & Sons, Inc.
8. Bell, D. (2005). Software Engineering for Students, a programming approach, 4th edition-2005.
Addison Wesley-Pearson Education Limited.
9. http://pc.net/helpcenter/answers/static_and_dynamic_web_pages
10. http://www.joomla.org/about-joomla.html
11. http://www.aviontechnology.net/advantages-and-disadvantages-of-joomla-cms- development/
80
Exploring the role of sentiments in identification of active
and influential bloggers
Mohammad Alghobiri1, Umer Ishfaq2, Hikmat Ullah Khan3 , Tahir Afzal Malik4
1,2
Department of Management Information Systems, Ibn Rushd College Management
Sciences, Abha, Kingdom of Saudi Arabia
3,4
Department of Computer Science, COMSATS Institute of Information Technology,
Attock, Pakistan
maalghobiri@kku.edu.sa1, tahir.malik@ibnrushd.edu.sa2, umer.bravo@gmail.com3,
hikmatullah@comsats.edu.pk4
Abstract. The social Web provides opportunities for the public to have social interactions and online
discussions. A large number of online users using the social web sites create a high volume of data.
This leads to the emergence of Big Data, which focuses on computational analysis of data to reveal
patterns, and associations relating to human interactions. Such analyses have vast applications in
various fields such as understanding human behaviors, studying culture influence, and promoting
online marketing. The blogs are one of the social web channels that offer a way to discuss various
topics. Finding the top bloggers has been a major research problem in the research domain of the
social web and big data. Various models and metrics have been proposed to find important blog users
in the blogosphere community. In this work, first find the sentiment of blog posts, then we find the
active and influential bloggers. Then, we compute various measures to explore the correlation
between the sentiment and active as well as bloggers who have impact on other bloggers in online
communities. Data computed using the real world blog data reveal that the sentiment is an important
factor and should be considered as a feature for finding top bloggers. Sentiment analysis helps to
understand how it affects human behaviors.
Keywords: Blogger, Sentiment, Social web, Big Data.
1. Introduction
The rise of Web 2.0 has turned online information consumers into information producers. Its
interactive and dynamic features allow the development of numerous innovative web services. The
Web 2.0 enables the masses to interact in social as well as collaborative environment also known as
the social networks. Blogs are one of widely used form of social networks, where users share their
views and experiences in the form of text, images or video forms [1]. Users get the advice of others
before making decisions, for example, choosing a place for shopping or buying products of a
particular brand. They listen and trust their opinions. In this way, they can be influenced by others
whom they are connected with and such users are termed as influential bloggers. Identification of
such users has been a real challenge [2].
Blogging has become a popular way of online interaction. People use it for voicing their opinions,
reporting news or mobilizing political campaigns [3]. The bloggers usually create their interest groups
and play a unique and significant role in social communities. The democratic nature of the
blogosphere has made it a lucrative platform for social activities and influence propagation [4].
Finding top bloggers has direct applications in various fields, including online marketing and ecommerce. By gaining the trust of influential bloggers, companies can turn them into their allies and
can save huge advertising sum [5]. Similarly, it helps in learning the market trends and to improve
insights for better custom care. Such top bloggers who have affected others can further assist in
reshaping the businesses and re-branding their products [6]. In this work, we explore the role of
sentiments and check how the sentiments of blog content correlate with the overall activities of the
top bloggers. We use the blog post content, compute their sentiment scores using SentiStrength, which
81
is the widely used technique to compute the sentiment of the content. We then find the correlation of
the sentiment with the characteristics of the top bloggers who are active as well as influential. In
addition to some standard features, we propose features related to sentiment. The blog of Engadget
has been used in this regard. In the following sections, first, we present the relevant work in section
2. Next, the proposed framework will be outlined where the research methodology is discussed in
details in section 3. The evaluation results are discussed in section 4. Finally, conclusion and direction
for further work will be presented in section 5.
2. Related Work
As blogging has gained considerable momentum in this era of time, people are highly influenced by
their fellow bloggers’ opinions. Researchers have highlighted various issues arising due to the
plethora of information piling up on the web. Social media is believed to be behind this revolution.
Such a development emphasizes the need of developing intelligent tools for mining valuable
information [7]. Therefore, social media contains huge amounts of information about people, their
customs and traditions. This valuable information can be exploited for better understanding
individuals and their communities [8]. Studies show that influence has multiple dimensions and types.
To differentiate one type from another and evaluate its importance over the other is real challenge [9].
PageRank is a well-known algorithm for ranking pages on the web. Zhou At el., [10] used PageRank
by taking into account nodes on the social media network for finding the opinions of leaders. The
authors in [11] identified the influential bloggers by ranking the blogs. Various blogs on the
blogosphere are compared and their relative importance is quantitatively measured. The work
proposed in [12] developed a data mining based model for identifying the leading bloggers. It
calculated the potential influence of influentials to maximize the spread of information over social
media. Most of the people on social media act as information consumers. Research presented in [2]
proposed a model which identifies the influential bloggers on the basis of information they produce
rather than their popularity within the social community.
A new ranking metric has been proposed in [13], which critically analyze well-known blogger ranking
algorithms, discussed their shortcomings. We found a standard model [14] which is based on four
basic blogger’s characteristics. It is novel in this regard that it identified influential bloggers. It
compared them with active ones and concluded that both are important. After that it was compared
with Pagerank as well and found that the model is better instead of Pagerank to find top bloggers.
This work is an extension effort by the same authors who actually initiated this domain of finding top
influential bloggers by introducing their model known as iIndex [15].
The most recent and comprehensive work that proposes a new metric, named as MIIB, finds the top
influential in a blogosphere community using modular approach. It proposes three modules and then
uses standard evaluation measures as well to show the significance of the proposed metric as well as
its modules [16]. The current work is continuation of earlier works [17] [18] on modeling in the
dynamic research domain of social web and semantic web.
3. Proposed Methodology
In this work, we propose the use sentiment feature. We use the existing features which are used to
compute a blogger’s activity as well as their effect within the community. The activity of bloggers is
calculated by counting their number of blog posts. It is an effective method which represents the
productivity of bloggers [4]. The recognition represents the position or status of a blogger in the social
community. The number of urls linking to the blog post (inlinks) and comments on a blog post is
taken as the recognition of the blogger. A large number of comments on a blog post show the interest
of other bloggers as they write their comments [15]. It is a good feature to calculate the impact of a
blogger on other bloggers as already considered by researchers in their works [4] [19] [15] [14].
Here, let us propose that adding a new feature of sentiment score, can increase the accuracy of the
model as discussed in detail in Section. 4. We examine how the sentiment score of a blogger correlates
with their overall activity as well as recognition. To the best of our knowledge, the sentiment feature
has not been used so far. Table. 1 shows the features included in the proposed model.
82
Table 1. List of Features used in the paper
Sr
1
2
3
4
Feature Name
The number of blog posts initiated by a logger
The number of URLs linking to a blog posts (inlinks)
The number of comments received in a blog post
The sentiment score of a blogger (computed using SentiStrength)
3.1 Data Source
We perform our experiments on data of Engadget8, a technology blog with real time features of a blog
site. The characteristics of the dataset are shown in table 1.
Table 2. Dataset Characteristics
Characteristics
Engadget
The number of Bloggers
93
The number of Posts
63,358
The number of Inlinks
319,880
The number of Comments
3,672,819
3.2 SentiStrength
We compute the sentiment expressed by a blogger in his/her blog posts. We use a standard method
for quantitative analysis of blog posts known as SentiStrength9. It is an opinion mining technique
which identifies the sentiment associated with a word and assigns positive and negative scores. For
that it uses a dictionary which categorizes the words into positive and negative just like a human
being. The snapshot of the online tool of the SentiStrength is given in Figure 1.
Figure. 3 Snapshot of SentiStrength v2.2 (Source: http://sentistrength.wlv.ac.uk)
3.3 Performance Evaluation Measures
Here, we discuss the both correlation measures that have been used for finding the correlation.
3.3.1 Spearman Correlation
8
9
http://www.engadget.com
http://sentistrength.wlv.ac.uk
83
A monotonic function is a function between ordered sets of two quantities. An increase in one quantity
causes the increase or decrease in the other. Spearman correlation10 uses monotonic function to
measure the association or strength of relationship between the two variables. It is computed using
the following equation:
𝜌 = 1−
6Σ𝑑𝑖2
(1)
𝑛(𝑛2 −1)
Where 𝑑𝑖 = 𝑥𝑖 − 𝑦𝑖 is the rank difference between the two ordered sets 𝑥 and 𝑦 which are
obtained by converting the variables 𝑋 and 𝑌 into ranks. 𝑛 represents the sample size of
the data.
3.3.2 Kendall Correlation
It is a non-parametric test to measure the dependence of two quantities upon one another. The
following equation is used to compute the dependence between the two quantities:
𝑛 −𝑛
𝜏=1 𝑐 𝑑
(2)
2
𝑛(𝑛−1)
Where 𝑛𝑐 and 𝑛𝑑 are number of concordant and number discordant pairs respectively.
4. Results Discussion
In this section, we examine how the sentiment score correlates with activity (number of posts) and
recognition (inlinks and comments) of a blogger. We perform this experiment on data of Engadget
blog. Table. 3 presents the results of our findings. The last three columns show the sentiment score,
activity and recognition score of the top-10 bloggers of Engadget dataset. As discussed earlier, the
blog posts of a blogger represent his/her activity. The number of inlinks and comments on a post
together constitute the recognition of a blogger. Sentiment score, on the other hand, is calculated using
SentiStrength, which calculates the sentiment score of content, the blog posts in this case, using a
mining based scaling system. A post has words with positive, negative, or neutral sentiment. These
words are analyzed on a scaling system and assigned positive and negative values. The system assigns
values from a dictionary which interprets the positivity or negativity with human level accuracy [20].
Table 3. Blogger ranking based on all the features
10
Rank
blog posts
Inlinks
Comments
Activity
Recognition
Sentiment
1
D. Murph
D. Murph
D. Murph
D. Murph
L. June
D. Murph
2
P. Rojas
T. Ricker
R. Block
R. Block
D. Murph
R. Block
3
R. Block
P. Miller
L. June
P. Miller
T. Ricker
P. Miller
4
P. Miller
N. Patel
J. Topolsky
P. Rojas
P. Miller
P. Rojas
5
D.
Melanson
R. Block
P. Miller
D.
Melanson
N. Patel
D. Melanson
6
T. Ricker
D.
Melanson
N. Patel
T. Ricker
J. Topolsky
T. Ricker
7
N. Patel
J. Topolsky
T. Ricker
N. Patel
R. Block
N. Patel
J. Topolsky
D. Melanson
E. Blass
8
E. Blass
C. Ziegler
D.
Melanson
9
J. Topolsky
R. Miller
C. Ziegler
C. Ziegler
C. Ziegler
J. Topolsky
10
C. Ziegler
V. Savov
P. Rojas
E. Blass
R. Miller
C. Ziegler
http://www.biostathandbook.com/spearman.html
84
Table 4. displays the correlation analysis of the proposed features with the two performance
evaluation measures of Spearman and Kendall correlation. The table shows the results of the features.
The comparison indicates higher values using Spearman equation. However, Kendall correlation has
smaller values for the comparing features. This is because Spearman’s correlation checks the
difference between the ranking orders of the two quantities whereas Kendall correlation analyzes all
the comparing quantities and tries to quantify the difference between the percentage of the concordant
pairs and the discordant pairs.
Table 4. The Correlation Analysis of the proposed features
Spearman Correlation
Kendall Correlation
Activity vs. Recognition
0.536
0.328
Activity vs. Sentiment
0.993
0.911
Recognition vs. Sentiment
0.485
0.401
Conclusion
In this paper, in addition to activity and influence, we also focus on the sentiment analysis of blog
posts. We use the blog post content, compute their sentiment scores using standard methods. Then we
correlate them with the characteristics of the top bloggers who are active as well as have influence on
other bloggers within the social network. Several features have been used and analysis has been
carried out on data of Engadget blog. The correlation between the sentiment and activity is very high
which show that the sentiment is an important factor for active bloggers but it is relatively low for
recognition part which suggest that sentiment does not have strong correlation with sentiment.
Sentiment is an important characteristic in social web and blog posts are not an exception. It helps in
analysis of overall community and human behaviors. For future work, we intend to include naming
disambiguation techniques proposed in [21] to check whether naming ambiguity plays role to find the
actual user or not. Other potential future work can be to find the top bloggers in a blogging community
using social network analysis metrics which are graph based measures.
References
1
2
3
4
5
6
7
Kale , A. Karandikar, P. Kolari, A. Java, T. Finin and A. Joshi, "Modeling trust and influence
in the blogosphere using link polarity," in International Conference on Weblogs and Social
Media (ICWSM), Boulder, 2007.
D. M. Romero, W. Galuba, S. Asur and B. A. Huberman, "Influence and passivity in social
media," in 20th international conference companion on World wide web, Hyderabad, 2011.
Johnston, B. Friedman and S. Peach, "Standpoint in Political Blogs: Voice, Authority, and
Issues," Women's Studies: An inter-disciplinary journal, vol. 40, no. 3, pp. 269-298, 2011.
N. Agarwal, D. Mahata and H. Liu, "Time- and Event-Driven Modeling of Blogger Influence,"
in Encyclopedia of Social Network Analysis and Mining (ESNAM), New York, Springer,
2014, pp. 2154-2165.
Sun and V. T. Ng, "Identifying influential users by their postings in social networks," in 3rd
international workshop on Modeling social media, Milwaukee, 2012.
G. Mishne and M. d. Rijke, "Deriving wishlists from blogs show us your blog, and we’ll tell
you what books to buy," in 15h International conference on World Wide Web, Edinburgh,
2006.
N. Agarwal and H. Liu, "Blogosphere: research issues, tools, and applications," ACM
SIGKDD Explorations Newsletter, vol. 10, no. 1, 2008.
85
8
9
10
11
12
13
14
15
16
17
18
19
20
21
S. Kumar , N. Agarwal , M. Lim and H. Liu, "Mapping socio-cultural dynamics in Indonesian
blogosphere," in Third International Conference on Computational Cultural Dynamics,
Washington DC, 2009.
J. Tang, J. Sun, C. Wang and Z. Yang, "Social influence analysis in large-scale networks," in
15th ACM SIGKDD international conference on Knowledge discovery and data mining, Paris,
2009.
H. Zhou, D. Zeng and C. Zhang, "Finding leaders from opinion networks," in IEEE
International Conference on Intelligence and Security Informatics (ISI), Dallas, 2009.
X. Song , Y. Chi, K. Hino and B. Tseng, "Identifying opinion leaders in the blogosphere," in
6th ACM conference on Conference on information and knowledge management, Lisbon,
2007.
Y. Singer, "How to win friends and influence people, truthfully: influence maximization
mechanisms for social networks," in Fifth ACM international conference on Web search and
data mining (WSDM), Seattle, 2012.
J. Bross, K. Richly, M. Kohnen and C. Meinel, "Identifying the top-dogs of the blogosphere,"
Social Network Analysis and Mining, vol. 2, no. 2, pp. 53-67, 2012.
N. Agarwal, H. Liu, L. Tang and P. Yu, "Identifying the influential bloggers in a community,"
in in: Proceedings of the International Conference on Web Search and Web Data Mining, New
York, 2008.
N. Agarwal, H. Liu, L. Tang and P. S. Yu, "Modeling Blogger Influence in a community,"
H. U. Khan, A. Daud and T. A. Malik, "MIIB: A Metric to Identify Top Influential Bloggers in
a Community," PLoS One, vol. 10, no. 9, pp. 1-15, 2015.
H. U. Khan and T. A. Malik, "Finding Resources from Middle of RDF Graph and at SubQuery Level in Suffix Array Based RDF Indexing Using RDQL Queries," International Journal
of Comuter Theory and Engineering, vol. 4, no. 3, pp. 369-372, 2012.
T. A. Malik, H. U. Khan and S. Sadiq, "Dynamic Time Table Generation Conforming
Constraints a Novel Approach," in International Conference on Computing and Information
Technology (ICCIT), Al-Madinah Al-Munawwarah, 2012.
L. Akritidis, D. Katsaros and P. Bozanis, "Identifying the Productive and Influential Bloggers
in a Community," IEEETransaction on System, Man, Cybernetics, Part C, vol. 41, no. 5, pp.
759 - 764, 2011.
M. Thelwall, K. Buckley, G. Paltoglou, D. Cai and A. Kappas, "Sentiment strength detection in
short informal text," Journal of the American Society for Information Science and Technology,
vol. 61 , no. 12, p. 2544–2558, 2010.
M. Shoaib, A. Daud and M. S. H. Khayal, "Improving Similarity Measures for Publications
with Special Focus on Author Name Disambiguation," Arabian Journal for Science and
Engineering, vol. 40, no. 6, pp. 1591-1605, 2015.
86
Morphological parsing of Albanian language: a different
approach to Albanian verbs
Ilir Çollaku1, Eşref Adalı2
Faculty of Computer Engineering and Informatics Istanbul Technical University,
ilir.collaku@itu.edu.tr1, adali@itu.edu.tr2
Abstract. The very first step when processing a natural language is creating a morphological parser.
Verbs in Albanian language are the most complex area of inflection. Besides irregular verbs, the ways
in which the regular verbs change their form while being inflected are hardly definable, and the
number of exceptions is huge.
In this paper, a different approach to Albanian verbs is made. Unlike traditional classification, based
on the inflection themes they take, verbs are classified into different verb groups. This way, the
inflection process looks clearer and more regular, as the affix remains the only changeable part of the
inflected verb. This way of approach, makes us able to process the Albanian verbs simpler and easier.
Keywords: NLP Albanian, morphological parsing, tagging, Albanian verbs, inflection theme
1. Introduction
Verbs in Albanian language are definitely the most complex among all parts of speech. Because of
their very rich entity of inflection forms, they are considered as the widest class of words and the
hardest one to fully master. Inflection forms in Albanian are built with changes in verb form and by
adding suffices, whereas the descent relationship between them is very complicated [1]. Verb
inflection forms, together with the construction of plural forms of nouns, are considered as two major
difficulties in Albanian morphology and its processing. Besides the political situation reigned in
Albania and Albanian-populated lands during last century, and the fact that no standard Albanian
language existed until the late 1970’s, the above mentioned difficulties can also be considered as a
reason why there is a lack of studies and research on NLP (natural language processing) Albanian.
Thus, Albanian language is considered as one of the most challenging languages to process [2].
2. Verbs in Albanian
Verbs in Albanian have their variable and constant forms. Variable forms of Albanian verbs have six
moods: indicative, subjunctive, conditional, admirative, optative and imperative; nine tenses: present,
imperfect, past simple, past, past perfect, (past) past perfect, future, future past and future front;
and two voices: active and passive. Constant forms of verbs are: past participle, infinitive, present
participle and negative. Regarding all these attributes, we can say that a single verb falls in 47 different
inflectional forms1. In almost every book of Albanian grammar, verbs are grouped based on their
inflection. According to traditional grouping, there are three groups of verbs: first inflection verbs
(verbs ending with -j in their singular 1st person indicative present tense active voice form, e.g. punoj
‘I work’), second inflection verbs (those ending in consonants, e.g. shoh ‘I see’), and third inflection
verbs (those ending in vowels, e.g. ha ‘I eat’).
2.1 Traditional grouping – a short analysis
First inflection verbs indeed have similar attributes. However, their inflections sometimes can differ,
e.g.
87
la.j - la.rë
mëso.j - mësua.r
‘I wash - to wash/washed’, but
‘I learn - to learn/learnt’
(note that the inflection theme of the second verb changes in its past participle, while that of the first
one remains same), and this can be useful in defining inflection rules when processing the language.
Second and third inflection verbs, on the other hand, differ from one another so much in inflection,
such that sometimes no similarities can be determined between verbs of the same group, e.g. the verb
pres ‘I wait’ changes its theme in past simple to prit.a ‘I waited’, while tremb, ‘I scare’ doesn’t change
it, tremb.a ‘I scared’, and no differences can be determined between verbs of different groups. A
second inflection verb inflects same in imperfect tense as one of the third inflection, e.g.
ha - ha.ja
hap - hap.ja
‘I eat - I was eating’ and
‘I open - I was opening’
Thus, when talking about morphological processing of Albanian verbs, in order to be able to define
inflection rules, they must be grouped in a more specific way.
2.2 A different grouping of Albanian verbs
A different method, but not unfamiliar to Albanian language, is grouping verbs based on their
inflection themes. When talking about verb theme, everyone thinks of its representative form, the
form it appears in dictionaries - in Albanian language it is the first person singular present tense
indicative mood active voice form of the verb (e.g.: punoj ‘I work’, laj ‘I wash’, ha ‘I eat’, marr ‘I
take’), but in morphology verb theme has another mean: it is the stem of the verb from which
inflection forms are built, and is called the inflection theme of the verb.
--1constant forms have been included, imperative mood has only the 3rd person, not all tenses are
applicable to all moods, and not every verb has its active and passive voice
For example, an inflection theme of the verb punoj ‘I work’ is puno., from which a myriad of forms
are built, like:
puno.j
puno.n
puno.jmë
puno.ja
...
‘I work’
‘you(sg) work/he,she works’
‘we work’ puno.ni
‘I was working’ puno.va
‘you(pl) work’ puno.jnë
‘I worked’
‘they work’
The same verb comes also with a different theme, that is punua.:
punua.r
punua.m
punua.kam
punua.kësha
...
‘working’
‘we worked’
‘I have been working (admirative)’
‘I had been working (admirative)’
Verbs with a single inflection theme aren’t few, e.g.:
fshi.
‘I clean’ la.j
‘I wash’ pi.
‘I drink’
But most of Albanian verbs have two or more inflection themes, and some of them can have up to
eight [1].
mëso.j
mësua.r
‘I learn’
‘to learn/learnt’
blua.j
‘I grind’ blo.va
‘I ground’ blu.het
‘it is ground’
88
shoh.
sheh.
‘I see’
‘you(sg) see’ shih.ni
dua.n
dash.ur
‘they want’ do.ja
‘to want/wanted’ du.hem
the.m
thua.
tha.shë
thë.në
thën.çi
‘I say’
‘you(sg) say’ tho.të
‘I said’
‘to say/said’
‘you(pl) say (optative)’
‘you(pl) see’ pa.shë
‘I saw’
‘I was wanting’ desh.a
‘I am wanted’
‘I wanted’
‘he/she says’ thu.het
‘it is said’
Fig. 1. Examples of verbs with two, three, four, five, and seven inflection themes. Based on the last
vowel of their inflection theme, the number of inflection themes, suffices they take and other structural
changes while being inflected, verbs in Albanian language are grouped in two major classes: verbs
inflection themes of which end with a vowel or a heap of vowels, and verbs inflection themes of which
end with a consonant. Further, depending on the number of inflection themes, each class is separated
to groups of verbs with one, two, three, up to eight inflection themes [1].
3. Parsing Albanian verbs – the solution proposed
This new method of grouping verbs based on their inflection themes, also changes the approach to
verbs in computational processing. Since Albanian language has extremely rich inflectional
paradigms and principles they follow aren’t systematic enough, they are not appropriate to be
presented by algorithms. Even if the necessary number of algorithms were designed for groups of
similar verbs, the number of exceptions would remain enormous.
3.1 Parsing based on the inflection themes
The number of verbs with similar structure but different inflectional specifications in Albanian
language isn’t small, e.g. verbs luaj ‘I play’ and bluaj ‘I grind’ are very similar, but the way they
inflect is different:
lua.j - luajt.a - luajt.ur
‘I play - I played - to play/played’, while
blua.j - blo.va - blua.r
‘I grind - I ground - to grind/ground’
Thus, the proposed solution which could make parsing Albanian verbs easier, is putting all their
inflection themes to the corpus. For example, auxiliary verbs kam ‘I have’ and jam ‘I am’ have the
greatest number of inflection themes, the first one has seven:
ka.m
‘I have’
ke.ni
‘you(pl) have’
kish.a
‘I had’
pat.a
‘I have had’
paç.im
‘we have (optative)’
pas.kam
‘I’ve had (admirative)’
ki.ni
‘you(pl) have (imperative)’
The other has eight:
ja.m
‘I am’ je.mi
‘we are’ është.
‘I was’
qe.shë
‘I have been’
qen.kam
‘I have been (admirative)’ qo.fsh
‘you(pl) are (imperative)’
‘he,she is’ ish.a
‘you(sg) be (optative)’ ji.ni
89
In this case, for two verbs kam and jam, all 15 records; ka, ke, kish, pat, paç, pas, ki, and ja, je, është,
ish, qe, qen, qo, ji should be put to the corpus. Thinking in similar way, plural forms of some nouns
should also be kept in the corpus (e.g. palë - palë ‘pair - pairs’, kalë - kuaj ‘horse - horses’, djalë djem ‘boy - boys’). As it can be seen from the above example, the three nouns are very similar to each
other (they all end up with ‘-alë’), but they construct their plural forms in completely different
ways.
3.2 Examples of verb inflection
Placing verb inflection themes on the corpus makes the definition of morphologic rules more likely
in computational environment. In contrast to inflection themes, suffices that the verbs take while being
inflected are more stable. They do not change at all, but only in certain cases, when depending on the
theme’s ending, an additional phonem (-j- or - n-) is placed between the theme and suffix. Suffices of
imperfect tense are constant for all verbs, they are added to the inflection form of the verb to build its
imperfect tense indicative and admirative forms, for example:
verb kam ‘I have’
Admirative form
‘I was having’
(inflection themes: ka., ke., kish., pat., paç., pas., ki.) Indicative form
kish.a
pas.kësh.- kish.im
kish.in
pas.kësh.a kish.e
pas.kësh.im kish.it
pas.kësh.in
verb jam ‘I am’
Admirative form
‘I was being’
‘I have been having’
pas.kësh.e kish.te
pas.kësh.it
(infl. themes: ja., je., është., ish., qe., qen., qo., ji.) Indicative form
‘I have been being’
ish.a
qen.kësh.- ish.im
ish.in
qen.kësh.a ish.e
qen.kësh.im ish.it
qen.kësh.in
qen.kësh.e ish.te
qen.kësh.it
verb mësoj ‘I learn’
Admirative form
‘I was learning’
(inflection themes: mëso., mësua.) Indicative form
mëso.(j)a
mësua.kësh.- mëso.(n)im
mësua.kësh.it mëso.(n)in
mësua.kësh.a mëso.(j)e
mësua.kësh.im mëso.(n)it
mësua.kësh.in
verb dal
‘I go out’
Admirative form
‘I was going out’
dil.(j)a
dal.kësh.- dil.(n)im
dil.(n)in
(inflection themes: dal., del., dil., dol.) Indicative form
‘I have been learning’
‘I have been going out’
dal.kësh.a dil.(j)e
dal.kësh.im dil.(n)it
dal.kësh.in
mësua.kësh.e mëso.(n)te
dal.kësh.e dil.te
dal.kësh.it
Fig. 2. Examples of verb inflection in imperfect tense, indicative and admirative forms.
90
Conclusion
Including inflection themes to the corpus enables us to handle inflection process of verbs more
systematically. Since in Albanian language the suffix is the part which keeps the information about
verb attributes (like mood, tense and voice), making verb themes static, reduces the scope of changes
in verb structure and eases inflection rules definition. Besides this, it also avoids definition of extra
inflection rules based on which the themes are built, and which, because of a lot of exceptions, still
would not offer a complete solution.
The only disadvantage here is keeping more than one record in the corpus for verbs which come with
several inflection themes, instead of keeping one record per verb. But again, this remains a better
solution than creating dozens of rules for each group of verbs, which also makes the processing last
longer.
References
1. Memushaj R., “Shqipja Standarde - Si ta flasim dhe ta shkruajmë”, Botimet Toena
2004, pg.87-94
2. Trommer J., Kallulli D., “A Morphological Tagger for Standard Albanian”
3. Piton O., Lagji K.,”Morphological study of Albanian words, and processing with
NooJ”, 2007
91
A Policeless traffic ticketing system with autonomous
vehicles
Mükremin Özkul and Ilir Çapuni
Dept. of Computer Engineering, Epoka University, Tirana Albania
Abstract. Besides being expensive, traffic violation monitoring systems rely heavily on a huge
infrastructure that incurs installation, operational, and maintenance costs.
Developing countries — where people do exhibit less traffic safety awareness — deployment of
such systems becomes a daunting task. A policemen cannot be everywhere, and a policeman can
also be bribed.
With the above goals in mind, in this paper we present an infrastructure- less and police-less traffic
violation detection system that relies solely on the broadcast messages between the vehicles and
secure communication of the vehicles with the transportation authority. Each vehicle should be
equipped with a small board box (or a smartphone) with a wifi antenna and 3G capability and
subscription. The system is highly scalable and can include pedestrians with smartphones. It is active
and operational wherever and whenever there are two participants in the range of each other.
Each participant has two roles to bear simultaneously. The first one is to report and prove its location
to the transportation authority. The second one is to report the presence of other vehicles in the
neighborhood to the transportation authority and flag those that disobey the traffic rules acting as
trustworthy and anonymous witnesses. This is a preliminary report on a still ongoing research project.
Keywords: Traffic violation system, traffic rules, faults, broadcast messages, V2V
1. Introduction
The traffic violation monitoring systems mostly rely on video detection based on real-time image
processing that use fixed roadside units, RSU, such as radars, cameras, or mobile service patrol
cars that is equipped with on board devices to identify the vehicles to report and to register
for issuing violation tickets. Unfortunately, correct operation of such systems is often conditioned
by good weather conditions. As a result are not efficient and accurate on rainy, snowy or in any
conditions that restrict visual contact between the vehicles on the road. Furthermore, image and
video computing algorithms are still costly and require a lot of computation power.
The coverage area of the video detection systems is limited by the visual coverage area. Because
of the high installation, maintenance, and costs of such cameras, deployment of these systems are
often limited to urban. In the US, most of the accidents resulting in bodily injures [5] occur
on the roads in the rural areas out of the scope of any traffic violation monitoring. Therefore, it
is important to have an intelligent monitoring system with a wide coverage without just relying on
such systems to enforce the traffic violations.
1.1
Our contribution
In this paper, we are introducing Traffic Violation Detection and Reporting System (TVDRS)
that
1. makes no use of policemen (which could be bribed).
2. is everywhere where there is traffic
3. needs no costly infrastructure.
In particular, we are interested in detecting, witnessing, and issuing a valid traffic violation ticket
for violations such as: speeding, moving on a wrong lane, parking on a forbidden place, driving in
the opposite direction on a one-way road, traffic light violation, not stopping on the pedestrian
92
crossways etc. To capture the traffic behavior with most of its realities, we use the model and
conventions presented in [1]. In short, in this model, a vehicle, a bike, a pedestrian. . . acts as
“automaton on the wheels” that has communication capabilities with their neighborhood of finite
size. From now on, we will call these objects vehicles, but with a clear intention that the definition
scales up to pedestrians and other participants as well. Therefore, as an automaton, its transition
function features the current traffic rules, whereas the local constraints (e.g. speed limit for a
certain part of the road at certain time intervals) is stored on a map which is loaded on the unit.
Each vehicle updates its state asynchronously. A violator is a vehicle that is not obeying to the
rules that regulate the traffic in the position that the vehicle is positioned. The violation is
detected and “witnessed” only by the vehicles in the space-time neighborhood of the violator
and is reported by them to a transportation authority (TA) that acts as a trusted party and is in
charge to collect violations and issue fines to the violators. A vehicle may be adversarial and may
not co-operate. Witnesses need to remain anonymous to the violator. We assume that vehicles
communicate with each other. Occasionally, a vehicle can communicate with the road side units
(RSUs) and use 3G/4G or later network if the topology of the road is covered by it. These
participants can communicate with the TA.
The desiderata of the system are as follows.
1. A violation ticket to a vehicle x for violation v that occurred at time t at position p
is generated only if a violation v at time t at position p was conducted by car x.
2. For a violation v of car x at time t at position p, at most one violation ticket
is generated.
3. The above holds, regardless of the behaviour of the participants or non- deterministic
nature of the environment.
4. Scales naturally so that other participants in the traffic (say, pedestrians or cyclists) can
participate.
1.2
Related work
A video-analysis system for real-time traffic-violation detection systems using cameras located
on the intersection in the urban areas is presented at [7]. The system model uses cameras with
content-analysis capabilities by using image recognition to classify objects in traffic in order to
detect and tract violations on the road such as one-way driving, and illegal bus-lane driving.
Radio Frequency Identification technology, RFID, enables automatic identification of vehicles.
Many counties in Europe use RFID tag systems in electronic toll collection, ETC, in order to
automatically charge passing vehicles at highway toll booths, or bridges. In [8], an architecture
of a traffic law enforcement system by using RFIDs is described.
1.3 The structure of the work
The rest of the paper is organized as follows. The following section starts with important –but
somewhat tedious – description of the underlying model that is given [1]. We recommend the
reader to skip the details on the first reading. Then, we proceed with giving the details of the
protocol.
2. Our System
2.1 The System Model
Our model uses a two-dimensional triangular grid whose sides are equal to l (say, l = 7.5 meters),
and we fix a coordinate system with two axes x and y as shown in Fig. 1. The main reason for this
choice is geometric: we can easily lay down such a grid on all kind of roads, regardless of their
→
−
−
curvature Fig. 2. For convenience we set set −
w = −→
x +→
y = (−1, 1) which gives us the set of
vectors
−
−
−
−
−
−
D = (→
x , −→
x→
y , −→
y ,→
w , −→
w)
93
which denotes the set of unit vectors in the vector space. A s ite in a grid is defined by a vector
→
−
p = (x, y).
An object is a pair
O = (s, P (O)),
where s is its (internal) state (from a finite set of states) and
P (O) = (x, y) ∈ Z2
determines its position on the grid, and it has 6 immediate neighbors in
−
−
−
IN (O) = {→
p +→
u |→
u ∈ D}.
A vehicle is an object that has its own propulsion and can move to one of the six immediate
neighboring sites.
Fig. 1. A vehicle placed at position P in a traffic lane bounded by a full boundary on the right,
and dashed on the left.
A part of the internal state of an object is its ID (say, based on a MAC address of the networking
device). Other information defining the state of an object may include current and maximum
allowed speed of the vehicle, direction, position of the vehicle on the road (in some other coordinate
system, say GPS), vehicle length, a priority index (say some vehicles like ambulance are given
higher priority), number and the identity of the passengers in the vehicle etc.
Fig. 2. A curve on the grid. A curve or and upheaval on the grid easily can be draw and
represented by using the triangular grid model.
There is a distinguished state called disabled to mark objects that cannot move or bond. We also
distinguish boundary objects that form boundaries. Boundary objects can be dashed or full. Any site
of the grid can hold at most one object that can span one or more consecutive sites at the same time,
except if one of them is of a dashed boundary type. In the illustrations, the boundaries of the roads
are marked as bold lines, like in Fig. 1. A site has its direction which is determined by the
direction of the boundary objects in its immediate vicinity. A lane is a list of consecutive sites
having the same direction between two boundaries. A vehicle object constantly updates its state by
applying a rule from the rule set. Additionally, the position of the same vehicle at time t may be
different from the time t − 1 after the rule application.
A motion rule of an object is a function
94
δ : (s, P (O)) 7→ (s0 , R)
−
→
−
→ →
−
where R = P (O) + u for some unit vector u = 0 .
The rule set prohibits a vehicle object from moving out of the road boundaries. Communication
and state synchronization of objects is defined as a bonding process. Each object can see the state of
objects of a constant size neighborhood around it. Together, two objects may form either a flexible
bond or a synchronization bond, or form no bond at all. However, the objects that only move in
the same direction can form bonds between each other. We can make use of non- vehicle objects
with a specific direction placed strategically on the grid to mark a road and its lanes, its boundaries,
its directions, maximum allowed speeds etc. The distance d(Vi , Vj ) denotes the Euclidean distance
between the two vehicles Vi and Vj . We extend this to define the distance between sites in a
natural way. System Evolution A configuration is a finite set of objects with bonds between
them. The switch from one configuration C1 to another one C2 is done by applying of a single
rule which defines motion of the vehicles as well. Whenever configuration C1 transitions to
configuration C2 using some rule r ∈ R, we will write C1 `R C2 . Within one unit of time, a
vehicle moves at least c1 cells and at most c2 cells, where c1 and c2 are two non-zero integer constants defined by the traffic rules, such that c1 < c2 . Whenever a configuration is obtained from a
previous by not applying an appropriate rule, we say that a violation occurred. For example, if a
site contains more than one vehicle at the same time, then a violation of a rule has occurred.
A trajectory is a finite sequence of configurations C1 , C2 , . . . , Cn , such that Ci ` Ci+1 for 1 ≤ i <
n. The whole system evolves as a continuous time Markov process, defined on a finite state space.
2.2 Procotol
The computation effort in our system is spread on the TA and on the vehicles using an optimal
ratio. We now describe the program of the vehicles and of the TA.
Fig. 3. Network
In the detecting part of the protocol, while beaconing, vehicles can detect that a vehicle is violating
the traffic rules. In order to witness that vehicle x has conducted a violation v at time t in position
p, a witness must prove that at time [t − τ, t + τ ] it was in the constant ball B(p, π), where τ
95
and π are two parameters whose value depends on the implementation. For this, vehicles need to
be able to prove their location to the TA, but in a way that does not intrude to its privacy: vehicles
should not know the IDs of the neighbors. The backbone of the system is the A-VIP protocol (see
[4]) devised to verify and infer positions of vehicles in vehicular network using anonymous
beaconing. The protocol is efficient (in terms of computational complexity) and robust against
common attacks. It also clearly identifies the cars that are not consistent in their proofs.
As we will see later, if we assume that a vehicle is equipped with a tamper resistant unit and
stipulate that not participating in the protocol is a violation itself, all the desired properties of the
system are fulfilled.
2.3 Vehicle rules
Now we describe the rules that run on each vehicle independently. Signup Each time a vehicle is
turned on, it will sign up to TA. This signup will ba valid for a certain amount of time and is
obtained by communicating with the TA over a secured channel using 3G/4G or RSU.
Suppose that a vehicle v sends a signup request to the TA at time t0 . The TA saves the request and
responds to it by sending the triplet (Kv , rv , ov ), where Kv is a short-term 128-bit AES
symmetric key, and rv and ov are two random integers.
Using these, both parties can compute a time dependent secret sv (t) and serves for the TA to
verify the identity of v and last time when it has issued a beacon. The sv (t) is computed as follows.
Both sides initialize a counter to the value rv and increment it by ov at every beacon. Then, the
value of the counter is encrypted using AES in counter mode.
Beaconing Once signed up, at every τb seconds, each vehicle broadcasts a beacon as follows.
Suppose that i is even. Then, the i-th beacon consists of the following.
1. Shared secret for the particular time instant sv (i).
2. Encrypted location information obtained by encrypting the string consisting of the
location padded with some flag bit1 zi−1 which is XOR’ed with the value (rv + iov ).
3. Plain location.
4. Current time.
Suppose that the is odd.
Now, we form a beacon using the shared secret sv (i − 1) from the previous time instant. The rest
is done the same as for even case.
Reporting Reporting is a second tasks performed by a vehicle. When a beacon is received, a vehicle
processes it using as detailed below.
Recall that a fault occurs whenever the transition between two configurations is not done according
to the transition function. For this, we define the OughtTo(v, sv , t) to be a set of possible states
that vehicle v can get to at time t + τb from state sv while obeying the rules. Now, for each beacon
that is received at some time instant, we compute its OughtTo. On the next beacon, we check if
the current state of the vehicle –
identified by the same shared secret (for two consecutive time instants), is in OughtTo. If not,
then clearly this vehicle has performed a violation of one of the rules.
Here are the details of the above. Suppose that a vehicle u receives a beacon of v.
On the next beacon, we compare if the state of the vehicle is in the OughtTo(x, sx , t). If not, then a
violation must have occurred.
Then, in a specific local table, vehicle u stores the beacon it received from v
along with
1. the time when the beacon was received
2. its own position when the beacon was received
3. an optional field Qu (v) that carries the signal power of v computed by u
1
The role of this bit will be clear later.
4. violation flag with the id of the type of the violation. If the violation flag is
0, then the violation id is null.
96
Furthermore, every τr seconds, u forms a message to TA containing all the recent beacons from its
neighbourhood. This message is sent through a secure channel after a successful authentication.
The very same beacon may be reported by many vehicles that are in the proximity of the vehicle
that generated it. Below we will show how the transcript of these reports will be combined to produce
a valid violation ticket.
Fig. 4. A Reporter with the local table
2.4
Transportation Authority
Whenever a TA receives a report, it processes them in order to (i) determine the location of the
vehicles that have announced their location through the beacons, (ii) verify the locations and the
violations, (iii) compute the actual position and violation flag of the vehicles that may have
advertised an incorrect location. Let S be the set of the positions and V be the set of vehicles that
need to be verified. Suppose that TA receives a report from vehicle u ∈ V . It then processes
entry-by-entry as follows.
1. read the time tuv when vehicle u has received the beacon from v,
2. for each w ∈ V computes
𝑖=𝑏
𝑡𝑢𝑤 − 𝑡𝑤0
𝜏𝑏
0
where 𝑡𝑤
is the time when TA has received the signup request from w;
3. using i, it retrieves the precomputed secret value xi that matches xi .
Suppose that such a match is found. Then, TA identifies v to be the vehicle that sent the beacon. It
extracts the quadruple associated to it in the report and performs the following operations.
1. Decrypt the location li that v has advertised via a beacon received by u at
time i and the flag z i−1 .
2. If the flag z i−1 = 1, the entry is discarded.
3. If the flag z i−1 = 0, the TA stores the position li and the position li in v v u
its own report table, together with the violation flag, violation ID, and the power indicator Qi (v)
if present. The role of z i−1 is to notify the TA that the i − 1-th beacon was affected by replay attack.
Now the TA performs the verification of position claims as in the plain vanilla A-VIP protocol given
in [4]. It is easy to see that our protocol inherits all the security properties of the aforementioned
protocol. Namely, the same arguments as for A-VIP show that our protocol is robust against various
attacks. However, we emphasize that violations can be detected only if there are cars witnessing it.
A special treatment is needed for the transmit-power attack and detecting a car moving on a wrong
direction. Namely, the protocol is devised to work assuming that all the vehicles do not maliciously
increase or decrease the transmission power. In such a case, a vehicle that is moving on a wrong way
(see Figure 5) may appear to be stationary to the vehicles behind. If there are no other vehicles in
front of the vehicles violating the one-way road in the figure, this violation remains undetected.
97
3. Speeding tickets
For the remaining of the paper, we focus on a specific fault of exceeding the maximum allowed
speed on a grid segment. A lane or more neighbouring lanes having the same speed limitations, and
directions are divided as a grid segments, roads, each with a unique identifier on the digital map
of a vehicle. A vehicle’s speed over a site is the ratio of the length of unit vector →−u , over
the time ti 7→ ti+1 .
Fig. 5. Power attack by a vehicle. V1 broadcasting with more power value than of the system
standard. Thus, it is observed by the other vehicles at a closer location on the grid.
Recall that when the vehicle V¯ is located inside one of the sites defined by OughtTo after applying
a movement rule, the vehicles is said to be obeying the traffic rules. Whenever the vehicle exceeds
the maximum speed boundary over a state update period of itself, that will cause the vehicle to
be inside a site location other than outside the OughtTo, than we a speed fault is committed by
the vehicle. The violations happen in an empty neighborhood is not monitored nor registered
because the lack of any reporters.
Fig. 6. The list of the witnessed violations delivered by a reporter vehicle to TA
98
Fig. 7. The grid map that is used for the simulation. The RSO is positioned near the destination
site.
First, the location information is processed and mapped on the vehicle map. Then, the state
information of speed, location, and lane of the violator vehicle is compared with the the site
properties information on the map. Whenever the information matches, the violation event is
validated as committed violation and a candidate ticket is issued and stored in the candidate
ticket list for the violator vehicle. If there exists a synchronization bond between the reporter and
the violator vehicles , the sync bond breaks and a flexible bonds forms. The bonding change
would prevent any vehicle in a cluster commit the same violation with the violator vehicle.
When a reporter vehicle enters in the communication distance of a RSO, the witnessed violations
is delivered to the TA and the reporter vehicle clears its list.
4. Simulation
In order to evaluate our proposed model, we have used OMNET++, Network Simulation, and
SUMO 0.22 an open source micro-traffic traffic simulator. VEINS 3.0 vehicular network simulation
framework establishes communication between OMNET++ and SUMO. OMNET implements the
IEEE 802.11p protocol stack at both the physical and the MAC layers. The state updates and
information exchange between the vehicles, i.e channel access and wireless communication is
implemented by using OMNET++ and VEINS. SUMO is used to simulate the movement updates,
and to obtain the environmental details of each individual vehicle on the grid. We use a portion
of the Tirana Durres highway, and a part of the urban area of Tirana as seen in Fig. 7.
Table 1. Simulation parameters for the simulation scenario
A RSO is positioned 100 meters away from the starting site at the beginning of the Tirana Durres
highway on Tirana side. The RSO transmits periodic information beacons to notify the reporter
vehicles and gather the local tables from the vehicles in the communication distance. A reporter
99
vehicle only delivers the information in the local list to the Trusted Authorities (RSOs). to ensure the
privacy of the vehicles. The length of the highway grid is approximately 8 km. and has two lanes,
and the speed violations take place in the highway section of the grid.
Fig. 8. The state of the local list of vehicle V0 during the simulation. The vehicles V2 and
V6 start speeding up and exceeds the maximal speed of the lane as they have already flex.
bonds to the V0 .
The vehicles of the same type assigned with different speeds of 80 km/h, 108 km/h, and 50 km/h. The
speed of a vehicle at a time t is obtained from OT . The first groups of the vehicles are capable of
reaching 80 km. and always maintain their speed under the inside the allowed speed boundary.
Together with the second group of vehicles with the speed of 50 km/h, the vehicles report the
vehicle’s violations. While the last group of the vehicles is capable of reaching 108 km/h, and
violates the maximum allowed speed on the highway segment of the simulation. In order to ensure
overtaking takes place, the violator vehicles enter into the simulation after the first mentioned
vehicle groups start the simulation. The number of vehicles on the grid is fixed to 1000, 100, and
200, respectively for each vehicle group. The trajectory is the same for all vehicles. At the
highway section of the map the maximum speed is bound at 90 km/h. Together with the
violator group of 100 km/h speed, this group forms a close cluster of vehicles before the
start of the highway to ensure to capture the first speed violation in the grid.
Fig. 9. The local lists are sent by the vehicles to the Road Side Object. After the
transmissions is completed the vehicles clean their local list.
Conclusion and Future Work
In order to be able to compute the OughtTo set, we use a non-sophisticated method of repeating
the same time dependent secret at two consecutive times. For the same reason, a vehicle releases
information about its precise location in plaintext. The first open question would be if there is such
a cryptographic tool that allows us to drop these –somewhat harsh – constraints. Second, a further
validation of the result should be made using a real implementation.
100
References
M. Ozkul, I. Capuni, ”An Autonomous Driving Framework With SelfConfigurable Vehicle Clusters” 2014 International Conference on Connected Ve- hicles
and Expo, pp.463-468,3-7 Nov. 2014 DOI 10.1109/ICCVE.2014.57
2
World Health Organization.World report on road traffic injury prevention, 2004.
3
SWOV Fact sheet. Headway times and road safety, December 2012.
4
Malandrino, et al., ”Verification and Inference of Positions in Vehicular Networks through
Anonymous Beaconing,” in Mobile Computing, IEEE Transactions on , vol.13, no.10,
pp.2415-2428, Oct. 2014 doi: 10.1109/TMC.2013.2297925
5
The Insurance Institute for Highway Safety (IIHS) General statistics.
6
http://www.iihs.org/iihs/topics/t/general-statistics/fatalityfacts/
7
state-by-state-overview#Rural-versus-urban
8
K. Nagel and M. Schreckenberg, “A cellular automaton model for freeway traffic,”
9
in J. Physique I, 2:2221 (1992).
10 Vijverberg, et al., ”High-Level Traffic-Violation Detection for Embedded Traffic Analysis,”
Acoustics, Speech and Signal Processing, 2007. ICASSP 2007. IEEE International
Conference on , vol.2, no., pp.II-793,II-796, 15-20 April 2007 doi:
10.1109/ICASSP.2007.366355
11 Vishnevsky, et al., ”Architecture of application platform for RFID-enabled traffic law
enforcement system,” Communication Technologies for Vehicles (Nets4Cars- Fall), 2014 7th
International Workshop on pp.45,49, 6-8 Oct. 2014.
1
101
Labeled-Image CAPTCHA: concept of a secured and
universally useful CAPTCHA
Mokter Hossain1, Ken Nguyen2, Muhammad Asadur Rahman2
1
University of Alabama, Tuscaloosa, AL 35487, U.S.A.,
Clayton State University, Morrow, GA 30260, U.S.A.
mokter@gmail.com1, {kennguyen, mrahman}@clayton.edu
2
Abstract. Captcha (Completely Automated Public Turing test to tell Computers and Humans Apart)
is a widely used online security tool that ensures that a computer program is not posing as a human
user. While smart programs with advanced image processing capability have already cracked picture
based captcha systems there is a need for making the test harder. This paper presents a design
prototype of a simplified type of labeled-image captcha where a picture of a common animal or
household item is marked with a number of different labels and the users will be asked to provide the
correct label for specific parts of the picture. Due to human’s familiarity with body shapes and part
names of such common pictures, they will easily identify a specific organ/parts of the picture. Such
labeled-image captcha tests are expected to be very easy for human users regardless of their culture,
age, gender, educational background and other discriminations but tough for the bots and automated
computer programs.
Keywords: Captcha, Turing test, labeled-image captcha, computer security.
1. Introduction
CAPTCHA stands for Completely Automated Public Turing Test to Tell Computers and Human
Apart, is a computer program that human can pass but the current computer programs cannot [1]. It
is mainly used to protect web-based services against automated computer programs. Fundamentally,
CAPTCHA is a cryptographic protocol whose principal mechanism is based on an Artificial
Intelligence (AI) problem called Turing test. Through a CAPTCHA test, users of a web service are
required to type in some distorted characters to authenticate that they are indeed a human. Such an
automated test is designed with a problem that a human user can solve easily but hardly possible for
current computer programs [2], [3].
A typical CAPTCHA is a group of words or images containing some distorted characters or other
images that appears somewhere, usually at the bottom of Web forms. Unlike the original Turing Test
in which a human person is employed as a judge, most modern-day CAPTCHAs are generated and
the judged by computers. The paradox of CAPTCHA is that a computer program that can generate
and grade tests that it itself cannot pass [2].
102
Fig. 3. A Typical use of a CAPTCHA in a website.
A typical use of a CAPTACHA test is shown in Figure 1 where some alphanumeric characters are
presented in distorted format. Once such a CAPTCHA is used in a Web-based service, a user of such
web service is asked to type in the alphanumeric characters in their exact order. If the user is successful
in doing so, s/he is considered as a human being and given access to the next steps of the web service.
There are a number of ways to implement CAPTCHAs in websites. However, the following three
types of CAPTCHAs are widely found in most websites [4]:
1. Text-based CAPTCHAs: This type of CAPTACHAs are based on a chunk of text images
presented in corrupted and distorted way to make them recognizable to human eyes but
unusual for pattern recognition programs.
2. Image-based CAPTCHAs: This type of CAPTCHAs are based on some kind of distorted
or faded images that require users to perform an image recognition task.
3. Audio-based CAPTCHAs: This type of CAPTCHAs are based on some audio or sound
items that require the users to resolve a speech recognition task.
At this age of free access to the social networking sites, bots and spams are becoming increasingly
annoying and unexpected threats to the Web-based application services and their users. Bots,
shortened form of robots, also called “intelligent agents” are automated or semi-automated program
tools that can accomplish some repetitive and routine tasks such as used for data mining or assisting
operations [13]. Bots can secretly move to another computer, especially through network or Internet
and then exploit that computer to launch some harmful activities or control attacks [14]. If
compromised, bots can simultaneously share resources across many criminal or unwanted operators.
They sometimes mislead human beings or legitimate users by trapping them to false actions or
requests. Thus, the use of CAPTCHAs is considered as an important aspect of modern computer and
Internet security systems. A complete list of CAPTCHA applications is maintained at
http://www.captcha.net. A human user should be able to answer the CAPTCHA in one to two tries.
This study is an initiative to study about how to make better CAPTCHAs that will still serve as robust
login security tool; however, will be easy to access and universally useful to all kind of web users.
Rest of the paper focuses on some fundamental and required information needed for this purpose.
2. Limitations and Criticisms of CAPTCHAs
Some common criticisms of human identification procedure used in CAPTCHAs is cumbersome and
time consuming which causes frustration and loss of productivity.
In text-based CAPTCHAs,
characters of the words are presented with multiple pictorial effects, and made an image with distorted
word so that the optical character recognition (OCR) machines or algorithms cannot detect the
characters. Although, this type of CAPTCHAs are most successful in establishing their goals, they
have gained dissatisfactory by many users, especially by visually impaired users [8].
103
Creation of text-based CAPTCHAs with multiple pictorial effects is costly and time consuming. A
more popular alternative is in the form of asking some questions such as “What is the sum of ten and
25?”, “What is 15 - five?", “What is the even digit in one thousand forty three?”, "Chest, brain, chin,
hair, and thumb: which is something each person has more than one of?”. The idea using text and
number together was to make simple CAPTCHAs that are accessible by users with visually impaired
[8]. There is a finite number of patterns used in this format making it is possible to design automatic
solver for this type of CAPTCHAs. Audio CAPTCHA is another alternative, however, it also has
limitations such as more time for users to solve the CAPTCHA and lower user’s success rate. Recent
research shows that 71% of audio CAPTCHA can be solve by computer programs, while human
success rate is about 70% [3]. Thus, audio CAPTCHA may not be as secured as once thought. A
comprehensive review of accessibility of different types of CAPTCHAs, especially for visually
impaired and elderly users, are seen in [8].
3. Security Issues with CAPTCHAs
3.1 Possible Security Holes in CAPTCHAs
A computer program that breaks the CAPTCHA security check can imitate as human and gain access
to services of the system provided by the website. For instance, it can automatically register with an
email server with a spam account, vote in online polls, or purchase tickets of a popular event to resale
later, etc. Thus, there is a trend among spammers to look for the security holes in CAPTCHAs, break
CAPTCHAs problems, and use them in performing many illegal activities that they could not do
otherwise. A weak CAPTCHA, regardless of it presentation format, is the one that can be recognized
by a computer program. Therefore, developing mechanisms to break CAPTCHA security such as
solving CAPTCHA challenges [9] helps identify and correct CAPTCHA weaknesses. Current
advancement in artificial intelligent enable computer programs to crack up to 90% of the CAPTCHAs
used in Google, Yahoo, and PayPal websites [10].
3.2 How CAPTCHAs are Broken
There are several approaches of breaking the identification process used in CAPTCHAs such as:
employing cheap human labors to decipher the distorted characters; exploiting program bugs used in
implementing CAPTCHA – this may allow attackers to bypass the CAPTCHA; using an existing
OCR software to detect the distorted characters; and so on. A common method to break CAPTCHA
security that use dictionary words is blind guessing a dictionary attack [5]. This blind guessing
technique yield a successful rate over 80% breaking Gimpy and EZ-Gimpy CAPTCHA security.
Another common method of breaking CAPTCHAs is using an image processing method. There are
three major steps in this method: pre-processing, segmentation, and character recognition [1]. First,
the color of distorted image of a CAPTCHA challenge is converted into a limited color (e.g. gray
scale) or binary formatted (black and white) image. Noises are removed and the image is segmented
into segments that may represent a character. Finally, in the character recognition phase the segmented
characters are deciphered using pattern-matching techniques [1]. Figure 2 shows typical architecture
of a CAPTCHA breaker system.
104
Input CAPTCHA image
Pre-processing
Segmentation
Feature Extraction
Character Recognition
Fig. 2. A typical CAPTCHA breaking architecture
A common practice that spammers and mass-ticket purchasers use is outsourcing CAPTCHA
problems solving to low-wage human workers in underdeveloped countries [7] The results, including
both the CAPTCHA problems along with their solutions, are save in a database for use in dictionary
type attacks.
4. Developing a Labeled-Image CAPTCHA
4.1 Needs for Creating New and Stronger CAPTCHAs
Character-embedded CAPTCHA is easy to generate and is also prone to be broken, Von Ahn’s group
proposed the audio CAPTCHA to take advantage of human’s ability to understand broken audio by
using context clues [3]. The audio CAPTCHA have a dual use: one for security purpose and the other
for transcribing old-time radio programs or historical speeches that are difficult for the current
automatic speech recognition systems.
Other interesting CAPTCHA researches such as grouping the animal types based on given pictures
[11] or rearranging randomly presented cartoon figures based on the meaning and/or utterances of
different cartoon panels [6].
Since CAPTCHA is a popular web-service security measure to deter unwanted automated access, a
more secure, easily recognizable by human and simple to create CAPTCHA is always a challenging
problem.
4.2 How to Create a New CAPTCHA?
CAPTCHAs are often programmatically created by combining distorted texts with irregular
background patterns to create an image [5]. The generated CAPTCHAs are more vulnerable when the
available stock images and background pattern are limited. Thus, most businesses rely on specialized
entity such as http://captchas.net for CAPTCHA services.
Figure 3 shows a simple Gimpy CAPTCHA – redesigned from the demo program scripts generated
form http://captchas.net written in Python. In order to integrate the distorted images of the CAPTCHA
tests both the user’s webserver and the CAPTCHA server (http://captchas.net, for instance) need and
share a common secret key. When the website requests for a CAPTCHA, it sends a random string to
the CAPTCHA server. The CAPTCHA server then calculates a password and returns the image in an
encrypted format and the solution of the image to the user website
105
Fig 3. A simple Gimpy CAPTCHA coded with Python
4.3 Concept of the Labeled-Image CAPTCHA
In this section, we propose an alternative version of image-based CAPTCHA that we named as
Labeled-Image CAPTCHA. In this type of Captcha a picture of an animal or an item will be presented
with a unique labels in each major organs of its body or structure and the user will be asked one or
more questions to identify a certain or some specific parts of the presented picture. For instance, the
whole body of a cat could be labeled as head, left ear, right ear, left front leg, right front leg, left rear
leg, right rear leg, body, tail, etc. Each of these parts could be labeled with different number or
character in random order, for instance, 1 for the left ear, 3 for the right ear, 4 for the tail, 5 for the
left front leg, 7 for the right front leg, 7 for the left rear leg, 6 for the right rear leg of the cat.
Figure 4 shows sample use of a labeled-image CAPTCHA where the user will be asked, for instance:
Which number is the closest label to the right front leg of the cat? As it is easy for human users to
easily detect such a common organ of a cat and enter the number 7 regardless of their age, race,
language and other distinctions they can easily solve such a Captcha problem.
Fig. 4. Possible use of a Labeled-Image CAPTCHA. The figure on the left is the enlarged labelCAPTCHA on the web-form on the left.
Design method of the form using a labeled-image CAPTCHA will be pretty similar to the design of
a simple image CAPTCHA, as shown in Figure 4. However, each of the labeled-images need to be
labeled carefully and the corresponding question to the labeled-image be written in a different field
with the database. Thus, the web registration form needs to have an image box for displaying the
labeled-image, a text label for displaying the associated CAPTCHA question and a text box for
inserting the CAPTCHA solution. With 26 characters in the English alphabet, 10 digit in the decimal
system, and at least 10 distinct parts on most of common objects and animal. This system would give
1036 different case-insensitive CAPTCHA patterns and 1062 patterns for case-sensitive CAPTCHAs.
An average person would have about 300,000 items at home [13] and is familiar with thousands of
animals, structures, and places. Thus, a database containing unique pictures of these subjects would
allow us to generate a very large set of unique CAPTCHAs that are trivial to human. Humans are very
good at recognizing patterns and we can easily recognize the same of subject from different placement
angle. Thus, the size of the picture database can be increased in several folds easily without hindering
human capability to recognize them by changing the angle, perspective, and projection of the subject
106
in the picture. The final number of distinct CAPTCHA patterns would be astronomic. Thus, making
it much harder for computers to crank through every possible combinations to find a match.
4.4 Possible Implication of the Labeled-Image CAPTCHA
Humans normally have enough knowledge and sense to be familiar with the different organs of a cat
or other animals no matter how large or different color of the animal is, but not the machines have.
Such kind of knowledge this type of CAPTCHA are intuitive for human but machines. A labeledimage seems to understood and interpreted uniquely by the users regardless of their culture, age,
gender, educational background and other discriminations. Due to abundant availability of pictures
and images of various animals, household items and many other common pictures it will be easy to
develop a large scale database for the labeled-image CAPTCHA. Recently Google announces a new
type of reCAPTCHA is able to identify a human user based on its mouse and cursor movement pattern
on the websites [12]. With the slogan “No more word puzzles: Google can tell you're human with one
click” Google claims that this technology will be more powerful and faster as a user will be asked
just to do a single click on a checkbox of the website using thus type of CAPTCHA. However, on
mobile devices, the new reCAPTCHA works a slight differently as there is no mouse or cursor
movement prior to tapping a button or place on a touch screen or mobile device [12]. So, instead of
clicking on a checkbox, users will be asked to select all or a number of images that correspond to a
specific clue in the CAPTCHA problem. Although, more details have not been known yet, the
proposed labeled-image CAPTCHA could be applicable for implementing Google’s new
reCAPTCHA for mobile devices. This seems to be a useful implication for the proposed labeledimage CAPTCHA.
Conclusion
CAPTCHA is a useful tool to deter unwanted machine automation, however, the traditional process
of CAPTCHAs generation make it easy for machine to overcome the intended security measure. In
this study, we proposed a new CAPTCHA prototype based on labelled-image, where a picture of a
common animal or household item will be marked with different labels. The users will be asked to
provide the correct label of specific part of the picture to unlock the service. Since humans are familiar
with the objects in the picture and the context of the question, they can easily decipher the problem.
Empirical study and additional research on these aspects could be the scope for further study in this
area. It is expected that further research on labeled-image CAPTCHA will introduce more secured
and useful impact on the digital society.
References
1. A. A. Chandavale, and A. M. Sapkal. “Algorithm for secured online authentication using
CAPTCHA,” pp. 292-297, 2010, IEEE
2. L. Von Ahn, M. Blum, and J. Langford. “Telling humans and computers apart automatically,” vol.
47, no. 2, pp. 56-60, 2004, ACM.
3. J. Tam, J. Simsa, S. Hyde, and L. V. Ahn. “Breaking audio captchas,” pp. 1625-1632, 2008
4. J. Yan, and A. S. El Ahmad. “Usability of CAPTCHAs or usability issues in CAPTCHA design,”
pp. 44-52, 2008, ACM
5. C. Pope, and K. Kaur. “Is it human or computer? Defending e-commerce with captchas,” vol. 7,
no. 2, pp. 43-49, 2005, IEEE
6. T. Yamamoto, T. Suzuki, and M. Nishigaki. “A proposal of four-panel cartoon CAPTCHA: The
Concept,” pp. 575-578, 2010, IEEE
7. M. Motoyama, K. Levchenko, C. Kanich, D. McCoy, G. M. Voelker, and S. Savage. “Re:
CAPTCHAs-Understanding CAPTCHA-Solving Services in an Economic Context.” vol. 10, pp.
4.1, 2010
8. S. Shirali-Shahreza, and M. H. Shirali-Shahreza. “Accessibility of CAPTCHA methods,” pp. 109110, 2011, ACM
107
9. J. Yan, and A. S. El Ahmad. “Breaking visual captchas with naive pattern recognition algorithms,”
pp. 279-291, 2007, IEEE
10. N. Summers. “Vicarious claims its AI software can crack up to 90% of CAPTCHAs offered by
Google, Yahoo and PayPal,” vol. 2014, no. October 25, pp. 3, 2013
11. B D. D'Souza, J. Matchuny, and R. Yampolskiy. “Zoo CAPTCHA: Telling Computers and
Humans Apart via Animal Image Classification,” 2014, Academy of Science and Engineering
(ASE), USA, ASE 2014
12. D. Hill. “No more word puzzles: Google can tell you're human with one click.” Available:
https://plus.google.com/ +DaveHill47/posts/8uG4LcxYQdi,” vol. 2014, no. 12/05, 2014
13. M. MacVean, “LATimes: For many people, gathering possessions is just the stuff of life.”
Available: http://articles.latimes.com/2014/mar/21/health/la-he-keeping-stuff-20140322.
108
Dual Tone Multi Frequency (DTMF) signal generation and
detection using MATLAB software
Nihat Pamuk1, Ziynet Pamuk2
1
Turkish Electricity Transmission Company
Sakarya University, Electric - Electronic Engineering Department
nihatpamuk@gmail.com1, ziynet@sakarya.edu.tr2
2
Abstract. In this study, Dual Tone Multi Frequency (DTMF) signal generation and detection is
implemented by using Goertzel Algorithm in MATLAB software. The DTMF signals are generated
by using Cool Edit Pro Version 2.0 program for DTMF tone detection. The DTMF signal generation
and detection algorithm are based on International Telecommunication Union (ITU)
recommendations. Frequency deviation, twist, energy and time duration tests are performed on the
DTMF signals. The algorithm recognizes the DTMF tones if they satisfy the recommendations,
otherwise they are rejected.
Keywords: DTMF, Signal Generation and Detection, Goertzel Algorithm, Frequency Deviation
1. Introduction
Dual Tone Multi Frequency (DTMF) is a system of signal tones used in the field of communication
whose application ranges from voice mail and help desks to telephone banking and controlling
robotics designs [1]. The DTMF signal is generated by the sum of two sinusoidal tones. One of them
is selected from a group of 697 Hz, 770 Hz, 852 Hz, 941 Hz named as low frequency group and the
second one is selected from a set of 1209 Hz, 1336 Hz, 1477 Hz, 1633 Hz called high frequency
group. By addition of two sinusoidal tones, four frequencies from each group gives a total of 16
combinations, which represented ten decimal digits, alphabet characters A, B, C, D and two special
characters “*” and “#” [2]. Table 1 shows the frequencies used for each signal.
Table 2. Frequencies used in forming DTMF tones.
Frequency
1209 Hz 1336 Hz 1477 Hz 1633 Hz
697 Hz
1
2
3
A
770 Hz
4
5
6
B
852 Hz
7
8
9
C
941 Hz
*
0
#
D
The signal tones should be detected by using Goertzel Algorithm. [3, 4, 5] have used Goertzel’s
algorithm for the detection of digital DTMF signals. Generating tones for all 16 digits (0 to 9, A to D,
* and #), the Cool-Edit Pro Version 2.0 program is used and the generating tones are used in
MATLAB software for detection. Automatic Gain Control (AGC) is added before detection which
provides the average output signal level is fed back to adjust the gain to an appropriate level for a
range of input signal levels [6]. DTMF detection is implemented by using Goertzel Algorithm and
tests are performed according to standards of International Telecommunication Union (ITU) by using
MATLAB software.
2. The Structure of Keypad and DTMF Signal
DTMF is a method of representing digits with tones for communications [7]. DTMF tones are used
by all touch tone phones to represent the digits on a touch tone keypad. DTMF signaling used for
109
many applications such as telephone banking, routing customer support calls, system control,
voicemail, and similar applications [8]. A DTMF signal represents one of sixteen touchtone symbols
(0-9, A-D, #, *) as shown in figure 1. Each symbol is represented by one of four frequencies in a low
frequency band and one of four frequencies in a higher frequency band.
Fig. 4. Touch tone telephone keypad
In figure 1, the symbols are shown in a matrix format. The columns are represented by frequencies in
a band between 1 kHz (kilohertz) and 2 kHz, and the rows are represented by frequencies in a band
between 500 Hz and 1 kHz. Whenever a key of a touchtone keypad is depressed, the DTMF is
generated by adding two signals and transmitted to device. The device that receives this dual tone
signal must detect which one of the four low frequencies and which one of the four high frequencies
has been received to determine which symbol is to be determined [9].
y (n)  A1 sin(
2 nf1
2 nf 2
)  A2 sin(
)
Fs
Fs
(1)
DTMF tones are generated like above equation 1, and f1, f2 frequencies denotes high or low frequency
tone of DTMF respectively. Fs represents sampling frequency.
3. DTMF Signal Generation and Detection
In communications, Discrete Fourier Transform (DFT) is used for frequency detection. DFT converts
the signal from time domain to frequency domain.
3.1 Goertzel Algorithm
Goertzel algorithm is used widely and it is most popular method in the worldwide. This algorithm
uses IIR filters that are tuned to eight DTMF frequencies. Direct realization of the goertzel algorithm
is showed in figure 2.
110
Fig. 2. Direct realization of the goertzel algorithm
Goertzel algorithm [10], can be seen from the equation 2.
Vk (n)  coeff (k).Vk (n1)  Vk (n  2)  x(n)
(2)
where Vk (-1) = 0, Vk (-2) = 0, x (n) represents input sample. Recursively compute for n = 0….N.
Compute once every N;
X (k)  Vk2 (N).Vk2 (N 1)  2.coeff(k).Vk (N).Vk (N 1)
(3)
The IIR filter coefficients for the row and column frequencies are calculated from;
fk
)
fs
(4)
2 fk
)
fs
(5)
coeff (k)  cos(2. .
The IIR filter coefficient for the second harmonics is found;
coeff _ 2nd (k)  cos(2. .
where WN-kN = e –j2пk
3.2 Validity Checking of DTMF Signal
Once all the row and column filter energies are calculated, a series of tests are executed to determine
the validity of tone and digit results [10], [11]; signal energy, twist, relative peaks, second harmonic,
duration check and frequency deviation.
Signal Energy: The DTMF detection algorithm uses adaptive threshold for the signal energy test.
The signal energy varies with the input signal level. Therefore it is difficult have a fix threshold for
this test. Fix threshold can be only used if the input level is same all the time. Automatic gain control
can be applied to the input.
Twist: Twist is the ratio of low frequency power high frequency power. If the row tone energy is
bigger than the column tone energy, this is known as normal twist. The threshold for the normal twist
is 8 dB (max). If the column tone energy is bigger than the row tone energy, the twist is described as
reverse twist and it can be 4 dB (max) [12], [13].
Relative Peaks: The energy of the strongest signal in each group (row and column tones) is compared
the energies of the rest of the tones in its group. The strongest tone must stand out from the other tones
in its group by a certain ratio [14].
Second Harmonic: The check of second harmonic provides whether the coming signal is DTMF or
speech. Speech has second harmonics, but DTMF signals do not have second harmonics.
Duration Check: According to ITU standards, the duration of DTMF signals should be 40 ms.
Frequency Deviation: The frequency deviation range should be between ± (1.5-3.5) %, where
accepting or rejecting the DTMF tone is up to the designer.
111
4. Experimental Work and Performance Test
In this section, first of all the DTMF tones should be generated to detect and test. After generating,
we start to detect these tones by using Goertzel Algorithm. We generate the DTMF tones by using
Cool Edit Pro program with different properties and characteristics. This program provides to save
the signals as a wav file format which is read wavered command in MATLAB program, so the transfer
is provided. The needed characteristics in Cool Edit Pro Version 2.0 are specified. The sampling rate
should be 8000 Hz because 8000 Hz is used in telephony applications.
The voice of human can be approximately 3200 Hz and according to the Nyquist criteria the sampling
rate should be twice frequency of the original signal, so 8000 Hz for sampling rate is enough to operate
by using human voice. If the sampling rate is bigger than 8000 Hz, it is meaningless and becomes
complicated and the execution time of program is getting longer. The channel is selected mono. Stereo
can be selected, but there is no need another channel for the application. The DTMF tones are
generated from Cool Edit Pro program like figure 3. If we write “7” in dial string part, you will obtain
this signal seen in figure 4. If we analyze the DTMF signal in frequency domain in Cool Edit Program,
we get the figure 5.
Fig. 3. The setting for dial string and time durations
Fig. 4. The DTMF signal “7”
112
Fig. 5. The frequency analysis of the DTMF signal “7”
In order to verify my MATLAB program, we get the DFT magnitude graph in MATLAB program
and saw that the program works correctly. Figure 6 shows the Goertzel DFT magnitude estimate of
the DTMF signal “7” tone. As can be seen from the figure 6, DTMF “7” tone contains 852 Hz and
1209 Hz sinusoidal. After the one digit is detected by program, we generated more than one DTMF
to be detected by taking buffer whose length is 128. We generated the all DTMF tones
“1234567890*#ABCD”.
Fig. 6. The DFT magnitude of the DTMF signal “7”
After that, we performed the following test; frequency deviation, twist, duration and relative peaks.
We performed the test two times to DTMF. After both of tests are satisfied, we decided whether or
not the DTMF is. After the DTMF is to be recognized twice, we said that DTMF is detected correctly.
The tests are performed by playing string “1234567890*#ABCD” five times. After the tests are
performed on DTMF signals five times, we detected the string “1234567890*#ABCD”. Therefore,
we show that DTMF algorithm works correctly. In second harmonic test, second harmonic energy is
found but second harmonic test is not performed because, we cannot talk off test tape. In the frequency
deviation test, when the frequency range is between ± 1.5 %, algorithm detected correctly. When the
range is selected more than ± 3%, the algorithm failed.
Conclusions
In this study, DTMF signal detection is implemented using MATLAB 7.0 program. The DTMF
signals are generated for each 16 digits in Cool Edit Pro Version 2.0 program. The signals are
strengthened by adding the algorithm of AGC. Then the DTMF detection is done by using Goertzel
Algorithm. For international communication some tests are to be done with the standard of ITU. First
of all the detection and tests are applied for one digit, but later the consecutive DTMF digits detection
113
is performed. The DTMF algorithm can be implemented in Simulink and the fixed point code can be
generated from Simulink and it can be used in real time telephony applications.
References
1. Singh, A.K.: Simulation of Dual Tone Multi Frequency Detection Using Bank of Filters,
International Journal of Engineering Research & Technology (IJERT), vol. 2, no. 6 (2013) 750755
2. Ravishankar, M.K., Hari, K.V.S.: Performance Analysis of Goertzel Algorithm Based Dual Tone
Multifrequency (DTMF) Detection Schemes, Departmental Technical Report (2004)
3. Mock, P.C.: Add DTMF Generation and Decoding to DSP-mP design, EDN (1985) 205-220
4. Gay, S.L., Hartung, J., Smith, G.L.: Algorithms for Multi-Channel DTMF Detection for the WE
DSP32 Family, International Conference on Acoustics Speech and Signal Processing (ICASSP)
(1989) 1134-1137
5. Bagchi, S., Mitra, S.K.: An Efficient Algorithm for DTMF Decoding Using the Subband NDFT,
International Symposium on Circuits and Systems, (1995) 1936-1939
6. Isaac Martinez, G.: Automatic Gain Control (AGC) Circuits Theory and Design, ECE 1352 Analog
Integrated Circuits-I (Term Paper), University of Toronto (2001) 1-25
7. AN218: DTMF Decoder Reference Design Application Notes, Silicon Laboratories Inc., 4635
Boston Lane, (2005) 1-36
8. Arslan, G., Evans, B.L., Sakarya, F.A., Pino, J.L.: Performance Evaluation and Real Time
Implementation of Subspace Adaptive and DFT Algorithms for Multi Tone Detection, Proc. IEEE
International Conference on Telecommunications, (1996) 884-887
9. US Patent 6608896.: Efficient Digital ITU Compliant Zero Buffering DTMF Detection Using the
Non-uniform Discrete Fourier Transform (2003)
10. Schmer, G.: DTMF Tone Generation and Detection: An Implementation Using the TMS320C54x,
Digital Signal Processing Solutions Application Note, Texas Instruments, SPRA096A (2000) 119
11. Mock, P.: Add DTMF Generation and Decoding to DSP-P Designs, DSP Applications with the
TMS320 Family, vol. 1, Texas Instruments (1989)
12. Edizkan, R.: DTMF Decoder Design for Fixed-Point DSPs (2005)
13. Emiya, V., Lucio, F.C., Vallot, P.D., Melles, D.: Generic Tone Dtection Using Teager-Kaiser
Energy Operations on the Starcore SC140 core (2004)
14. Proakis, J.G., Manolakis, D.G.: DTMF Tone Generation and Detection: An Implementation Using
the TMS320C54x, Digital Signal Processing Solutions Application Report, Texas Instruments,
SPRA096 (1997) 1-56
114
Randomizing Ensemble-based approaches for Outlier
Lediona Nishani1 , Marenglen Biba2
1,2
Department of Computer Science University of New York Tirana, Albania
ledionanishani@gmail.com1, marenglenbiba@unyt.edu.al2
Abstract. The data size is increasing dramatically every day, therefore, it has emerged the need of
detecting abnormal behaviors, which can harm seriously our systems. Outlier detection refers to the
process of identifying outlying activities, which diverge from the remaining group of data.
This process, an integral part of data mining field, has experienced recently a substantial interest from
the data mining community. An outlying activity or an outlier refers to a data point, which
significantly deviates and appears to be inconsistent compared to other data members. Ensemblebased outlier detection is a line of research employed in order to reduce the model dependence from
datasets or data locality by raising the robustness of the data mining procedures. The key principle of
an ensemble approach is using the combination of individual detection results, which do not contain
the same list of outliers in order to come up with a consensus finding. In this paper, we propose a
novel strategy of constructing randomized ensemble outlier detection. This approach is an extension
of the heuristic greedy ensemble construction previously built by the research community. We will
focus on the core components of constructing an ensemble –based algorithm for outlier detection. The
randomization will be performed by intervening into the pseudo code of greedy ensemble and
implementing randomization in the respective java code through the ELKI data-mining platform. The
key purpose of our approach is to improve the greedy ensemble and to overcome its local maxima
problem. In order to induce diversity, it is performed randomization by initializing the search with a
random outlier detector from the pool of detectors. Finally, the paper provides strong insights
regarding the ongoing work of our randomized ensemble-based approach for outlier detection.
Empirical results indicate that due to inducing diversity by employing various outlier detection
algorithms, the randomized ensemble approach performs better than using only one outlier detector.
Keywords: outlier detection, ensemble outlier detection, greedy ensemble, randomized ensemble,
ELKI
1. Introduction
The exponential growth of large databases has led to the need for monitoring, examining and
predicting economical, weather forecast or other various procedures in the whole world. In these
processes not so often occur rare events, distinguished from the daily basis processes, which can
harm or can deteriorate the respective process. These rare behaviors are called outlier or
anomalies, which can happen very infrequently. The most popular definition of outlier is “an
observation, which deviates so much from other observations as to arouse suspicions that it was
generated by a different mechanism” [1]. Outlier detection has found a substantial attention
from many areas of research respectively in disclosing malicious transactions in the
banking system, intrusions detection in the networking system and privacy data in the healthcare
databases. Data mining and machine learning algorithms have emerged in order to solve the outlier
detection tasks. Data mining communities have categorized outlier detection methods by four
different groups [2]: statistical reasoning, distance-based, density-based and modeled- based
approaches. In regards to statistical reasoning [3], the data points are represented as a
stochastic distribution where outliers are identified based on the relationship they have with the
stochastic distribution of the data points. Statistical reasoning approaches undergo through various
limitations in higher dimensionality data because they can face significant difficulty in computing
the multidimensional distribution of data in these conditions. On the other hand, distance-based
approaches proposed in [4], [5], [6], [7], [8], [9], are intended to enhance and mitigate the
115
shortcomings that statistical approaches pose to outlier detection problems. Their key idea is based
on estimating distances between data points and assigning score to data. The event, which has the
larger score, is pronounced as outlier. Regarding density- based approaches, various pieces of
work are presented, too: [10], [11], [12], [13]. They rely on computing the densities of local
neighborhoods. Referring to model- based techniques, they determine the normal behavior by
making use of predictive models just like neural networks or unsupervised support vector machines.
Making use of just one algorithm such as a density-based or a distance-based algorithm does
not summarize the entire kinds of outliers because there are methods, which are good at detecting
some kind of outlier and others, which perform better in other domains. Therefore, in order to
provide the whole truth, it is appropriate to integrate different outlier detection outcomes by
means of an ensemble-based approach coming up with a consensus finding. Ensemble analysis is
a kind of method, which aims to reduce the model dependence from the specific dataset or data
locality. Usually ensemble approaches are referred as the combination of the data outcomes
executed in independent ways. Ensemble-based approaches for outlier detection are more difficult
to be employed compared to classification or clustering domain due to the combination of small
sample space and the unsupervised nature. This is why the state of-the-art of ensemble analysis for
outlier detection has been neglected from the research community and has not been profoundly
analyzed in depth. Ensemble for outlier detection has inherited from the ensemble-based
classification two major properties of constructing an ensemble named: accuracy (to be better
than random) and diversity (to perform different errors in different algorithms) in order to enhance
the detection process. Diversity is referred to overcome making the same errors due to the correlated
output results. Accuracy consists of having the high performance detection rate and low false
detection rates. In the remainder of the paper, related background will provide motivations of the
ensemble-based analysis and the most popular and significant line of research from the first
attempt to the latest works. By randomizing this ensemble algorithm we intend to enhance the
detection rate and to mitigate the false positive rates. After disclosing the motivation that underlies
this line of research, the next section deals with the proposed strategy that we will motivate our
research by stating the reason why we have chosen to follow this idea. Finally, in concluding
of the paper we reasonably indicate some future open issues that we intend to deal with in the future
work and fill the gap of the previous researches.
2. Related Work
In this section, we are going to describe some works, which have been extensively cited and have
opened new line of researches. Ensemble analysis has raised attention for the first time in the
classification problems; in the supervised learning techniques. Extensive research was undertaken
in building ensembles from single classifier algorithms in order to enhance effectiveness. This
line of research has a sound theoretical background [15], [16], [17], [18], [19]. Besides
classification, ensemble analysis has found a considerable attention in constructing ensembleclustering methods [2]. Referring to the field of outlier detection, there are some strong attempts
in the implementation of some outlier detection algorithms. Therefore, this domain has been called
an “emerging area” [21]. Usually ensemble methods are considered as meta methods [22] because
they process the output of their own methods. One procedure that can induce diversity is dubbed
as bagging [14]. This is an ordinary process employed in the classification and clustering
techniques. However, this kind of technique has not been so successful and has not found sound
theoretical framework for implementation. Instead, various techniques have been presented that
tackle the score and the ranking problem. In [14] was proposed a breadth-first traversal through
the outlier rankings in order to combine algorithms. Then it is performed the comparability of
the retrieved scores. Calibration approaches aimed to fit outlier scores that have been detectors
outcome have been converted into probability estimates [23]. In [24] was proposed an algorithm,
which has generated some scores centered on their mean and been scaled by their standard
deviations.. On the other hand, statistical reasoning has been carried out to make sense of
different outlier scores into converting through outlier probabilities [25]. In this paper has been
disclosed the likelihood of improvement across combining various approaches, but it lacks
on applying of measure of current diversity of correlation among algorithms members. Ensemble
116
analysis has been deployed successfully in outlier detection for high dimensional data where
multiple subspaces are examined in order to identify outliers [26], [12], [27], etc.
Outlier detection methods are categorized in two major groups based on the kind of outlier they
intend to discover: global methods and local methods. The distanced- based definition of outlier
was proposed for the first time from [28]. It comprises the first data based-oriented approach in the
domain of outlier detection. On the other hand, local methods are been extensively explored
from significant strategies. Variants of local approaches have been proposed in [13] and [29].
Another work related to detecting the principle of local outlier is LDOF [9]. Rather than in terms of
speed and computation time, ensemble based approaches exhibit the capability of enhancing the
performance of their algorithms in terms of quality of detection. Shubert et al. [2] referred
to similarity measure in order to accurately interpret different outlier rankings and to evaluate the
diversity that different outlier detection algorithms generate. Not all this line of research proposed in
the body of knowledge has pursued the inducement of diversity from novel methods. Subsequently,
they have not found out new tools of inducing diversity for the selected pool of outlier detector. That
is why is advocated that the theoretical fundamentals of ensemble based approaches for
outlier detection is not advanced and immature even though they have borrowed some sound
principles from the rich tradition of supervised leaning techniques for ensemble analysis.
In [30] is introduced subsampling technique as a meaningful mean of inducing diversity between
detector members. It is given evidence that executing an ensemble of various algorithms detectors is
a subsample of the data is more effective than the other methods of inducing diversity. The
paper major contributions is in demonstrating empirically that it can be constructed an ensemble
for outlier detection, which outperforms the individual outlier algorithms if they are executed
individually.
3. Proposed Strategy
Our strategy derives from the greedy ensemble of Schubert et.al [33]. We have modified this
ensemble algorithm by increasing its diversity throughout randomization. The key principle of the
greedy ensemble relies on the diversity maximazation and at the same time keeping the size of
the ensemble small. It is demonstrated [21] that using the diversity factor can enhance and migliorate
the performance of the ensemble substantially. The available outlier detectors utilized as ensemble
members are variants of kNN-based outlier detection: Local Outlier Factor (LOF) [10], LDOF [9],
LoOP [13], KNN [7], k-NN [4], [32]. Authors have carried out experiments using the ELKI data
mining platform. After exploring and examining in depth the greedy approach, we have inspected a
crucial gap in this line of research that let us to do further improvement. The key idea is that while
greedy approach is iterating in order to find the best outlier detector and thus discarding with no future
chance to evaluate the algorithms back, we propose to perform a randomization. In this moment of
time, for instance, the ensemble must not choose the best of algorithm closer to the target vector
or the less closer, but one random detector. We had this idea due to the logic that the best outlier
sometimes does not lead to the best outcome. It happens that while searching for the best result in
local climbing search, it may find out just the local maximum data point, but not the global maximum
of the whole dataset. Therefore, in order to escape the local maxima, we need to employ
randomization techniques to the greedy ensemble. Moreover, by randomizing we make sure that
the search will continue in a random data point. We have conceptualized that this kind of
methodology can lead to substantial increase of accuracy and diversity. Randomization in practice
will be induced by modifying the pseudo code and at the same time, implementing this change in
the java code of the greedy ensemble construction, which is provided from the data-mining group.
We have used the Netbeans java interface in order to modify the code. In java, the class of random
represents randomization. This randomization problem is analogues with the rolling dies problem.
We will set a probability ɑ, which will be given externally, and according to this probability are
going to be select or to choose randomly the remaining outlier detectors. After selecting a random
algorithm, we will test and run the greedy ensemble with the new added detector. Empirical results
show that the outputs will be diversified and new kind of outliers not discovered before will
be captured from the randomized greedy ensemble approach. Therefore, we can foster that
deploying randomization techniques in the greedy ensemble will contribute in identifying and
capturing new malicious data points.
117
Conclusion and Future work
This paper is a short overview of the major steps that the construction of the randomized
ensemble-based approach will follow and under what circumstances our research direction is
founded. The key idea of our work is that while greedy approach is iterating in order to find the best
outlier detector and thus discarding with no future chance to evaluate the algorithms back, we
propose to perform a randomization. In this moment of time, the ensemble must not choose the best
of algorithm closer to the target vector, but one random detector. Randomization in practice will be
induced by modifying the pseudo code and at the same time, implementing this change in the java
code of the greedy ensemble construction, which is provided from the data-mining group. We have
used the Netbeans java interface in order to modify the code. In future, we plan to construct an
ensemble based on the combining and selected different outlier algorithms with different
parameters. Numerous experiments are predicted to be carried out. Through various experiments,
we are going to build our approach by providing strong experimental results.
References
1. Hawkins, D.: Identification of Outliers. s.l. Chapman and Hall, (1980)
2. Schubert, E., Wojdanowski, R., Zimek, A., Kriegel, H.: On Evaluation of Outlier Rankings and
Outlier Scores.. Anaheim,
CA : SIAM, Proceedings of the 2012 SIAM International
Conference on Data Mining. (2012) pp. 1047-1058
3. Hadi, A. S., Rahmatullah Imon, A., H.M. Werner, M. :Detection of outliers. WIREs Comp.(
2009) pp. 57–70
4. Angiulli , F. and Pizzuti, C.: Fast outlier detection in high dimensional spaces. Proc. PKDD
(2002) pp. 15–26
5. Knorr, E. M., Ng, R. T., Tucanov, V.: Distance based outliers: Algorithms and applications.
VLDB (2000) pp. 237–253
6.Orair, G. H., Teixeira, C., Wang, Y., Meira Jr, W.; Parthasarathy, S.: Distance-based outlier
detection:Consolidation and renewed bearing. PVLDB (2010) pp. 1469–1480
7. Ramaswamy, S., Rastogi, R. and Shim, K.: Efficient algorithms for mining outliers from
large data sets. Proc. SIGMOD (2000) pp. 427–438
8. Vu , N. H. and Gopalkrishnan, V.: Efficient pruning schemes for distance-based outlier detection.
Proc.ECML PKDD (2009) pp. 160–175
9. Zhang, K., Hutter, M. and Jin, H.: A new local distance based outlier detection approach for
scattered real world. Proc. PAKDD (2009) pp. 813–822
10.Breunig, M. M., Kriegel, H. P., Ng, R., Sander, J.: LOF: Identifying density-based local outliers.
In Proc.SIGMOD (2000) pp. 93-104
11. De Vries, T., Chawla, S. and Houle, M. E.: Finding local anomalies in very high
dimensional space. Proc.ICDM (2010) pp. 128–137
12. Keller, F.; Müller, E. and Böhm, K.: HiCS: high contrast subspaces for density-based outlier
ranking. Proc.ICDE (2012)
13. Kriegel, H. -P., Kroger, P., Schubert, E., Zimek, A.: LoOP: local outlier probabilities. Proc.
CIKM (2009) pp. 1649-1652
14. Lazarevic, A. and Kumar, V.: Feature Bagging for Outlier Detection. Proc. KDD. ACM.
Chicago, Illinois (2005) pp. 157-166
15. Breiman, L.: Bagging Predictors. Machine Learning (1996) pp. 123-140
16. Chawla, N., et al.: SMOTEBoost: Improving the Prediction of Minority Class in Boosting.
In Proceedings of the Principles of Knowledge Discovery in Databases PKDD (2003)
17. Freund, Y. and Schapire, R.: Experiments with a New Boosting Algorithm. In Proceedings of the
13th International Conference on Machine Learning.Bari : s.n. (1996) pp. 325-332
18. Joshi, M.; Agarwal , R.; Kumar, V.: Predicting Rare Classes: Can Boosting Make Any Weak
Learner Strong? Proceedings of the Eight ACM Conference ACM SIGKDD International Conference
on Knowledge Discovery and Data Mining.
Edmonton, Canada: (2002)
19. Kong, E., Dietterich, T.: Error-Correcting Output Coding Corrects Bias and Variance. In
118
Proceedings of the 12th International Conference on Machine Learning. San-Francisco CA : (1995)
pp. 313-321
20. Ghosh , J., Acharya, A.: Cluster ensembles . Wiley Interdisciplinary Reviews: Data Mining and
Knowledge Discovery. (2011) pp. 305–315
21. Schubert, E. Generalized and Effiicient Outlier Detection for Spatial, Temporal, and HighDimensional Data Mining. Munchen, Germany : s.n. (2013)
22. Aggarwal, Ch. C.: Outlier Ensembles. ACM SIGKDD Explorations Newsletter I2012) pp.
49-58
23. Gao, J. and Tan, P. -N.: Converting Output Scores from Outlier Detection Algorithms into
Probability Estimates Hong-Kong : IEEE Data Mining ICDM '06. Sixth International
Conference (2006) pp. 212 - 221
24. Nguyen, H. V., Ang, H. H. and Gopalkrishnan, V.: Mining outliers with ensemble of
heterogeneous detectors on random subspaces. Proc. DASFAA (2010) pp. 368-383
25. H.-P. Kriegel, P. Kroger, E. Schubert, and A. Zimek.: Interpreting and unifying outlier scores.
Proc.SDM (2011) pp. 13-24
26. He, Z., Deng , S., Xu, X.: A Unified Subspace Outlier Ensemble Framework for Outlier
Detection. Advances in Web Age Information Management (2005)
27. Muller, E.; Schiffer, M.; Seidl, T.: Statistical Selection of Relevant Subspace Projections for
Outlier Ranking. ICDE Conference. (2011) pp. 434–445
28. Knorr , E. M. and Ng, R. T.: A unified notion of outliers: Properties and computation. Proc.
KDD (1997) pp. 219-222
29. Papadimitriou, S.; Kitagawa, H.; Gibbons, P. B.: LOCI: Fast Outlier Detection Using the Local
Correlation Integral. IEEE 19th International Conference on Data Engineering (ICDE'03) (2003)
30. Zimek, A.; Gaudet, M.; Campello, R. J. G. B.; Sander, J.: Subsampling for Efficient and Effective
Unsupervised Outlier Detection Ensembles. Proceeding KDD '13 Proceedings of the 19th ACM
SIGKDD international conference on Knowledge discovery and data mining. ACM. New York, USA
(2013) pp. 428-436
31. Zimek, A., Campello, R. J. G. B. and Sander, J.: Ensembles for Unsupervised Outlier
Detection. ACM SIGKDD Explorations Newsletter (2013) pp. 11-22
32. Kriegel, H. -P.; Schubert, M. and Zimek, A.: Angle-based outlier detection in highdimensional data. KDD (2008) pp. 444–452
33. Shubert, E., Wojdanowski, R., Zimek, A., Kriegel, H.-P.: On evaluation of outlier rankings and
outlier scores. 11th SIAM International Conference on Data Mining (SDM) s.n., (2011)
34. Zhang, K.; Hutter, M.; Jin, H.: A new local distance-based outlier detection approach for
scattered. Proc. PAKDD (2009) pp. 813-822.
35. Papadimitriou, S., et al: Fast outlier detection using the local correlation integral. Proc.
ICDE (2003) pp. 315-326
36. Ghosh , J., and Acharya, A.: Cluster ensembles. WIREs DMKD (2011) pp. 305–315
37. Nguyen, H. V.; Ang, H. H.and Gopalkrishnan, V.: Mining outliers with ensemble of
heterogeneous detectors on random subspaces. In Proc. DASFAA (2010) pp. 368–383.
119
10 W DC-DC converter based LED driver circuit design
Ö. Faruk Farsakoğlu1, H. Yusuf Hasirci1, İbrahim Çelik1
Kilis 7 Aralık University, Department of Electrical and Electronic Engineering, Kilis,
79000
ffarsakoglu@kilis.edu.tr
1
Abstract. Considering the total amount of energy consumed in the world, energy used in lighting is
of utmost importance. For this reason, systems used in lighting should be energy efficient, and mıre
efficient lighting elements should be preferred. LED, as a solid-state lighting system, is more energy
efficient than the lighting systems with conventional lighting elements. Some of the LEDs used in
solid-state lighting systems are named as standard 5mm LEDs and power LEDs. Because power LEDs
have more light efficiency than standard 5 mm LEDs, they are produced as an alternative to
conventional light sources. Power LEDs draw much more current than standard 5 mm LEDs. LEDs
need LED drivers that provide them with constant current to run efficiently, and have a long life. The
present research studies 10 W DC-DC converter based current limited LED driver circuits.
Simulations were created for these LED driver circuits, and they are analysed through their
simulations. Efficiency, input current, total circuit loss, output current and power values are measured.
In this context, output current and efficiency values of the driver circuits are analysed in terms of
energy efficiency, and optimised in accordance with energy efficiency.
Keywords: Current limited LED driver, energy efficiency, power LED, DC-DC converter
1. Introduction
Considering that, one fifth of the energy consumed in the world is used for lighting, saving of the
energy used in lighting systems, and increasing the efficiency of this energy becomes inevitable [1].
The most important factor in energy saving is energy conservation. The elements used in lighting
(light source, secondary elements etc.) should be chosen among systems running correctly and
efficiently for a better saving, and more efficient use of energy. Solid-state Lighting (SSL) systems
are can conserve energy more than conventional lighting systems. Additionally, SSL systems can be
an alternative technology to develop lighting systems. Fig. 1 presents the efficiency of some LEDs
and conventional lighting systems. Solid-state lighting technology is based on the use of light emitting
diodes (LED) [2]. The first solid-state lamps were used in display LED applications [3].
120
Fig.0. Efficiency values for some solid-state LEDs [4]
Semi-conductive LED lighting elements are produces as power LEDs as a better alternative to
conventional lighting elements such as incandescent lamp, fluorescent tubes, and halogen lamps due
to their many advantages such as higher energy efficiency, longer life (50.000 – 100.000 hours),
needing less maintenance, smaller equipment size, and therefore smaller size, variety in light and
colour, endurance to impact and vibrations, and less energy consumption. Power LEDs are commonly
used in lighting sector due their high lumen values. A schematic diagram of LEDs is presented in Fig.
2. Current values of power LEDs are much higher than standard 5mm LEDs. If these LEDs are to be
used, they need a special type of driver circuit.
Fig. 2. Schematic diagram of power LEDs [5]
2. LED Driver Circuits
LEDs need driver circuits as they can only run on DC voltage polarity, and need current limiting [6].
LEDs can be divided into two groups as; current limited and voltage limited. Current limited LED
driver circuits are generally preferred, as overcurrent results in heating in LEDs. Various current
driving methods are studied in terms of energy efficiency. Some of these are; LED driver circuits with
LM317 regulator circuit, simple driver circuits with resistance, and DC-DC convertor based LED
driver circuits. Driver circuits with LM317 integrated circuit can be used within 1.25V and 37V input
voltage range, and up to 1.5A output current [7]. LM317 is not suitable for high power LED drivers.
As presented in Figure 5, output voltage is kept constant with resistance attached to adjust ends.
Around 4-5 W of energy is lost in the LM317 used in this driver, when the circuit is in transmission.
This decreased the efficiency of afore mentioned driver circuit. In the simple driver circuit with
resistance, 0.7 A of current is transmitted when energy is first given to the circuit. In time, LEDs start
to heat, and forward voltage of LEDs start to decrease. The current transmitted on the circuit increases
in this case, and when the heat reaches to a certain extent, current applied to LED increases to 0.75
A. Increased current results in more heating, and LED can be broken in time without necessary
precautions.
2.1 DC - DC Convertor Based LED Driver Circuit
10 W DC-DC convertor based current limited LED driver circuit was designed using LM3429
dropper-amplifier regulator. Fig. 3 presents a 10 W DC-DC convertor based current limited LED
driver circuit. This regulator can run up to 75 V input voltage, and comply with dropper, amplifier,
dropper-amplifier circuit topologies.
121
Fig.3. 10 W DC - DC convertor based LED driver circuit
Efficiency (%)
Fig. 4, 5 and 6 present efficiency, output power, total energy loss, input current values, and changes
in LED current for 10 W LED driver circuit. Fig. 4 presents the change in efficiency of 10W DC-DC
convertor based LED driver circuit according to input voltage. Accordingly, efficiency increases
when input voltage is lower.
90
89
88
87
86
85
84
83
82
Efficiency
16
20
24
28
32
36
Input Power (V)
Fig. 4. Change in efficiency of 10W DC-DC convertor based LED driver circuit according
to input voltage.
Change in output power and total loss of 10W DC-DC convertor based LED driver circuit according
to input voltage is presented in Fig. 5. Accordingly, output power is constant in 9.30W value, and
total loss increases as input voltage is higher.
122
2.00
15
1.80
13
1.60
11
1.40
9
1.20
7
1.00
5
16
20
24
28
32
Output Power (W)
Total Loss (W)
Total Loss
Output Power
36
Input Voltage (V)
Fig. 5. Change in output power and total loss of 10W DC-DC convertor based LED driver circuit
according to input voltage
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1
0.8
0.6
0.4
0.2
LED Current (A)
Input Current (A)
Change in input and output current of 10W DC-DC convertor based LED driver circuit according to
input voltage is presented in Fig. 6. Accordingly, as input voltage value increases, input current
decreases, and LED current doesn’t change.
Input Current
LED Current
0
16
20
24
28
32
Input Voltage (V)
36
Fig.6. Change in input and output current of 10W DC-DC convertor based LED driver circuit
according to input voltage
Conclusion
DC-DC convertor (buck) based current limited LED driver circuit was designed as 10W, and
simulations were created. Input current, LED current (output current), efficiency, output power, and
total loss values were measured for DC-DC convertor (buck) based current limited LED driver circuit.
It was observed that, as input voltage increased, efficiency decreased accordingly. It was also found
that output power didn’t change according to input voltage in DC-DC convertor (buck) based current
limited LED driver circuit. Energy spent by LEDs didn’t change according to input voltage, while
total loss increased. As input voltage increased, LED current increased as well, while input current of
the circuit decreased. It was found that, 10W DC-DC convertor (buck) based current limited LED
drivers were more efficient than simple driver circuits with one resistance, and LED driver circuit
with LM117 regulator.
123
References
1
2
3
4
5
6
7
Enerji ve Tabii Kaynaklar Bakanlığı, Kamuda Verimli Aydınlatmaya Geçiş, Ocak 2009,
Ankara, http://www.enerji.gov.tr/File/? path = ROOT%2F1%2FDocuments%2 FBelge%2FK
VAG Raporu. pdf,
Perpifiai, X., Werkhoven, R. J., Vellvehi, M., Jordai, X., Kunen, J. M., Jakovenk, J., Bancken
P. and Bolt, P.: LED Driver Thermal Design Considerations for SolidStateIEEE, 978-1-46731513-51121, (2012) 1-5,
Arik, M., Petrosk J., and Weavery, S. Thermal Challanges In The Future Generatıon Solıd
State Lıghtıng Applıcatıons, IEEE, 0-7803-7152-6/02 (2002 ), 113-120
Dupuis R. D. and Krames,M. R, History, Development and Applications of High-Brightness
Visible Light-Emitting Diodes,Journal of Light wave Technology,Vol. 26, (2008)1154-1171
PowerLED nedir?, http://www.ledportali.com/powerled-nedir/ (2015)
Gürbüz, Y. Güç Faktörü Düzeltmeli Güç LED Sürücü Tasarımı ve Gerçekleştirilmesi, Selçuk
Üniversitesi Fen Bilimleri Enstitüsü, Konya (2012)
LM317 Datasheet, http://www.ti.com/lit/ds/symlink/lm317.pdf
124
The influence of cloud computing and mobile technology in
our activities
Ramadan Cikaqi1 , Ilir Morina2
STS “Skender Luarasi”
cikaqiramadan@hotmail.com1, ilirimorina@gmail.com2
1,2
Abstract. Through this paper we present the development of information technology especially in
cloud computing and mobile technology as an integral part of everyday life, as well as the advantages
that cloud computing and mobile technology offers to the business community and the private one.
Furthermore, it presents the possibilities offered by this technology on the data preservation compared
with the traditional ones.
Due to the increased demand for big space, it was necessary to establish a centralized data storage
which has brought with it more innovation and advantage, in comparison with other media for data
storage in personal memories. This opportunity has influenced individuals, companies and the
community in general, because of the access to data at any time and from almost every country.
Essential principle includes achieving increased productivity benefits at different scopes including
ecological impact, reducing managerial costs, investments in infrastructure and exploitation of
common resources.
Keywords: Cloud Computing, Mobile Technology, Storage, Data.
1. Introduction
“Cloud Computing” is a platform with an impact in the world of information technology, which brings
on itself a lot of possibilities in the computing field. This platform is being developed on fast paces
and is being increasingly used by both the service operators and their clients. The development of
such a nature of “Cloud Computing” is enabled by the development of new computing technology
which enables the use of computing infrastructure in a completely different form and its operation in
a reasonable cost. “Cloud Computing” includes sending of computing recourses, which have the
ability of further amplify according to demands of clients without a need on big investments in the IT
Infrastructure as preparation on new implementations [1].
There are a lot of variations on defining the “Cloud Computing”, but some of the most appropriate
ones are: National American Institute of Standards and Technology (NIST) “Cloud Computing”, is a
model which enables suitable access and based to the needs in a sharing group of computing
configuration recourses. For example, computer networking, servers, storages, applications and other
services, can be easily set or retrieved on a minimal managerial or services operators intervention.
Whilst, according to Gartner IT glossary we have – “A computing model where the elastic and
amplified resources of the IT are offered as services to a lot of clients using internet technology”
[2][3].
2. The usage of cloud computing and mobile technology
Except the basic usage which cloud computing has offered, which have already shown their results,
these usages are being increased over time on both more services and possibilities within cloud. The
base usage remaining are the online access to data, access at any time, access anywhere covered by
internet, finding of tools online, online storage and avoidance of any insecurity threatening the
hardware equipment, access to data by two mobile devices at the same time, the usage of contact
results, calendar, call list and apps data at one account, access to services, automatic savings, usage
of free memory space on cheap cost, secure closing of devices by cloud, the separation of data and
their usage by other users [4].
125
The utilization of Cloud Computing and Mobile Technology abroad is at a satisfactory level, but how
is that in our country, as well as balance between the private and public community is very interesting
to follow, because the mobile technology is very present, starting from equipment such as: Desktop
PC, I Phone, android, blackberry, is always updated on both sectors, while the application of Cloud
Computing on these technological equipment is insufficient or better to describe it as being very low.
In our country, the cloud as a platform is little incorporated and it is mainly used privately, and this
platform is used, more or less at certain businesses, while in institutions and especially in schools has
started to be used due to the reason that a lot of branches are professional and the usage of cloud is
mandatory because the tasks and students homework are uploaded online and are presented at the
classroom by the online access. In order to present an actual example of the usage of cloud in one of
our country’s institutions, we have shown in the graph the usage of cloud by students, who had an
assignment which they set on cloud, at the school where we work, thus we have the following results:
100%
80%
60%
Class X
40%
Class Y
20%
Class Z
0%
Year 2012
Year 2013
Year 2014
Figure 1 The comparison of the usage of throughout the years
This is conducted out of 5 given assignments for the number of students on a n academic year and
this can be presented in the form a task and can be in graph presented as follows:
Assignment: For Grade X with 22 students during the years 2012 – 2014, are given 5 tasks uploaded
in cloud on a school year, thus we have:
22 ∙ 5 = 110
(1)
Which shows 100%
18+11+21+16+14 = 80/110 =0,72
(2)
126
Which as an amount is calculated in Excel to be presented in percentage, while in the math’s form we
have:
110------100%
(3)
80---------X%
80∙100 = 110∙X
8000/ 110
X = 72%
3. Opportunities of using the cloud computing and mobile technology
If cloud is considered as an opportunity, then this opportunity is given to all the people who have
access in at least one of the cloud providers, whilst as a technology includes trainings and the
advantages that every individual has on information technology in order to access this technology and
to gain out of it the benefits offered. Moreover, it can use cloud as a more convenient platform of data
setting in a storage offered by this platform as well as the security of the data. As a result, how do we
understand and use it, is by connection of a lot of ways into a single one, which in cooperation with
mobile technology offers a suitable services to data to proper destinations and to gaining a space
storage easily managed.
The user opportunities of this platform are powerful and amplified to secure unlimited resources to
storage, in which the information have a great security. Furthermore, it is easier for the user because
they can have access at every time and everywhere and they can share with others. Concerning cloud
computing clear trend is centralization of computing resources in large data centers. In addition, it has
changed the concept of storages, reliability and easy access on data, compared to previous years and
traditional storages, which has used the hardware which has been threatened at any time by different
factors. Among the best known providers of the usage of the possibilities of cloud space are, such as:
Microsoft Windows Azure, Google App Engine, Amazon EC2, IBM Smart Cloud.
4. Advantages of Cloud Computing and Mobile Technology
During the years of cloud practicing, it is noticed that there have been both the advantages and
disadvantages of this platform, hence, through this letter we have presented the advantages above at
the usage of cloud, thus, among the most emphasized ones are:
Pay-Per-Use Model / Only Pay for What You Utilize - Unlike many computing programs where the
package comes with unnecessary applications, the cloud allows users to literally get what they pay
for. This scalability allows for you to simply purchase the applications and data storage you really
need. "Pay-Per-Use" Billing Model Cloud usage policy defines that you will be billed for cloud
resources as you use them. This pay-as-you-go model means usage is metered and you pay only for
what you consume. Users have to pay only for the resources they use, ultimately helping them keep
their costs down. Because this pay-for-what-you-use model resembles the way electricity, fuel and
water are consumed, it’s sometimes referred to as utility computing.
Mobility - Users can access information wherever they are, rather than being dependent on the
infrastructure.Elasticity - The cloud is elastic, meaning resources, software and the infrastructure can
be scaled up or decreased, depending upon the need. Service Based Usage Model - Availability of
large computing infrastructure and the services on need basis.
Mobility - One of the main advantages of working in the cloud is that it allows users the mobility
necessary in this day and age of global marketing. For example, a busy executive on a business trip
in Japan may need to know what is going on at the company headquarters in Australia. Rather than
having to make phone calls and rely on the reports of employees, they can simply check updated
statistics online. The Internet is, for the most part, everywhere. Therefore, cloud computing allows
the mobility necessary for success.
Versatile Compatibility - It is an ongoing debate: which is better, the Mac or PC? Despite which side
of the fence you stand on this argument, it makes no difference when it comes to implementing cloud
solutions into a business model. Users are often surprised to find that the various cloud apps available
are accessible on both platforms.
127
Individuality - One of the most convenient aspects of working in the cloud is that it is compatible with
aspects specific to the company. For example, cloud IT services can be scaled to meet changing
system demands within a single company [5].
Other advantages are big storages, back up on cloud as a service to retrieve data, automatic
synchronization of devices, as well as power savings, which is done by third parties, but which plays
an important role, in a certain country, can be an example in environmental protection.
5. Disadvantages of Using the Cloud Computing and Mobile Technology
Except the good opportunities offered by Cloud Computing and Mobile Technology they often have
their flaws as well as difficulties that are carried by these technologies for their cooperation between
them. The most emphasized disadvantages are:
Privacy: data remain in the device out of the company’s structure, what obliges the latest to trust the
service provider in cloud for its confidentiality, as well as the data often remain at the ex-employee,
thus, it causes the possibility on misusing the data by the third parties by unauthorised access.
Security: The data are always online and can be target of some criminal computing activity.
Continuation of the Service: Data remain physically in one place, which usually is away from
company’s facilities, and if there is not any connection to internet then the data are inaccessible.
Downtime: This is the panic situation for business owner, when site goes offline for some time. No
doubt that with this issue has to face everybody. Even Amazon, Google and Apple websites face these
problems. So think about, what your business has compared to those big companies. There was no
other option for complete solution to avoid downtime completely [6].
Transferring big data: transferring data or files of large size will not be feasible, as it will take both
time and money.
6. The usage of Cloud Computing in business and private sectors
Cloud business, or the implementation of cloud in business sector is presented in a growing trend due
to some reasons which bring sufficient benefits for the companies, among which we can distinguish:

Accessible Service 24 hours a day: data and other business application remain in the
devices out of the company’s structure and administration, which can be accessed
wherever you are and whenever you need.

Lowering the risk: the insurance of the data is guaranteed by the third party, service
provider in cloud. Herein, we understand the security against the unauthorized access,
within or outside the company, e.g., fire, device malfunction, etc.

Low Support Cost: herein, it is included the cost which deals with installation, the usage
and the updating of the needed applications on the functioning of the IT infrastructure, we
include the operative systems, antiviruses, firewalls, etc.

Low maintenance cost: the company receives the cloud service, thus it does not need to
undergo into maintenance cost for the physical devices, though, because this is a service
provider company’s responsibility.

Low managing cost: the company does not have to undergo the software license
procurement or on the replacement and providing with new physical equipment. Instead,
they only have to pay the monthly fee on usage of the network at the whereabouts of the
storage of the data. IT Infrastructure is sold by the Re provider as a service on which the
businesses pay only for what they get. (Pay On Demand).
Cloud private, the strategy of the implementation of Private ‘Cloud’, can be compared with the
traditional strategy of services, nonetheless, this implementation method uses technology which are
used by ‘Cloud Computing’, such as virtualization to secure advantages for individuals. ‘Cloud’
private, virtualization technology in order to build the infrastructure of technology of an individual
by offering high advantages compared to other infrastructures. In our country, the situation is almost
completely opposite in terms of the usage of these two systems, cloud private is used more than the
cloud business. Regarding the mobile technology, it noted that it is present at every individual, thus,
being the strongest point compared to the business one in the usage of cloud.
128
6.1 The comparison of the usage of Cloud on Business and Private Sector in Theranda
As stated above, strangely the same situation appears in the town of Theranda, where the usage of
cloud private is more than the usage of cloud business and according to the questionnaire we have the
following ratio: in the business sector the usage is 14% of the cases, while in the private sector it
consist of 18% of the usage.
Ratio of the use of private cloud and business cloud
Business
14%
Private
86%
Figure 2 Ratio of the use of private cloud and business cloud
7. The advantages and disadvantages of storage of data in traditional
memory devices in ratio to cloud
There are a lot of advantages of storing of data in cloud in comparison to traditional memory devices,
such as: physical security, which is no threat compared to traditional ones, which can be lost, broken
or stolen; data security is higher; larger storage space; law protection storage space; a part of the
storage space is given to the user without charge; accessibility everywhere and at any time; there is
no need to carry the memory space with you, you will have it only if there is access to network, etc.
Apart from these advantages of cloud, there are also some disadvantages, in comparison to the
traditional memory devices, starting from the inaccessibility to data in case there is no internet
connection, the losing of order in uploads if there is interruption on the network, upload and download
are variable and depend on network bandwidth.
Conclusion
Cloud computing and Mobile Technology have influence the current society in a way how the services
are offered by them, as well as these services have influenced the society in various ways, by making
it accessible both as an individual and as a business. These accesses into two forms has also reviled
the differences between them, therefore, we can conclude that cloud as a platform, in our country, is
used more individually than in a business sector.
All the advantages of cloud, in comparison to the traditional memory devices, such as: USB-s, Harddiscs and other memory devices have challenged the traditional ones, therefore, it has been found that
the aforementioned point, cloud remains in a much better situation, even though there are still a lot of
hesitations to its usage due to the fear on its security, though, there is nothing to fear of in this regard.
Moreover, if we compare cloud to the traditional memory devices, the latest can be easily lost or
stolen, thus, we will have the same result. All in all, cloud remains in a higher and better position, in
regard to its storage capacities, management and a lot of other above mentioned advantages.
At the different institution, such as schools, especially those technical ones, there is an increasing
tendency of cloud usage, based on our study results at school classes presented on the graph.
Facing with the issue of current fear on cloud usage and professional preparation to recognize cloud
by clients will be the main challenge in the future, nevertheless, the usage of cloud is constantly in
growing and in scalability of new opportunities within itself.
129
References
1
M. Armbrust, A. Fox, R. Griffith, A. Joseph, R. Katz, A. Konwinski, G. Lee, D. Patterson, A.
Rabkin, I. Stoica, and M. Zaharia, “Above the Clouds:A Berkeley view of Cloud Computing,”
University of California, Berkeley, 2009. [Online]. Available:
2
http://d1smfj0g31qzek.cloudfront.net/abovetheclouds.pdf
3
Peter Mell, Timothy Grance, The NIST Definition of Cloud Computing, Recommendations of
the National Institute of Standards and Technology, September 2011.[Online]. Available:
http://csrc.nist.gov/publications/nistpubs/800-145/SP800-145.pdf
4
Gartner, Stamford, U.S.A. Offical site Available: http://www.gartner.com/it-glossary/cloudcomputing
5
Dimosthenis Kyriazis, Cloud Computing Service Level Agreements, Exploitation of Research
Results, Brussels, June 2013.
6
Fernando Macias, Greg Thomas, Cloud Computing Advantages, Cisco, [Online]. Available:
http://www.cisco.com/web/strategy/docs/c11-687784_cloud_omputing_wp.pdf
7
Anca Apostu, Florina Pucian, Study on advantages and disadvantages of Cloud Computing
University of Bucharest Offical site, [Online]. Available:
8
http://www.wseas.us/e-library/conferences/2013/Morioka/DSAC/DSAC-16.pdf and official
site http://cloudcomputingadvices.com/cloud-computing-advantages-disadvantages/
9
Larry Coyne, Shivaramakrishnan Gopalakrishnan, John Sing, Ibm Private, Public, and Hybrid
Cloud Storage Solutions, July 2014, [Online]. Available:
10 http://www.redbooks.ibm.com/redpapers/pdfs/redp4873.pdf
11 Chris Harding, Cloud Computing for business, [Online]. Available:
12 http://www.opengroup.org/sites/default/files/contentimages/Press/Excerpts/first_30_pages.pdf
130
Bounty techniques for web vulnerability scanning
Tanzer Abazi1, Mentor Hoxhaj1, Edmond Hajrizi1, Gazmend Krasniqi1
1
UBT, Computer Science,
{tanzer.abazi, mentor.hoxhaj, ehajrizi, gazmend.krasniqi}@ubt-uni.net
Abstract. With the advancement of technology and the raising of massive amount of data, the used
techniques for data security are continually a challenge. This paper contributes on identifying the gaps
and evaluating security level on web portals hosted or managed by Republic of Kosovo institutions
or businesses, whose data privacy and security could be a big business concern. The results have been
obtained from real case scenario, where 25 security researchers have contributed in white hack
activities. These activities, were part of a one day conference called. “The DAY when hacking is
legal”, held in Pristine
Keywords: bugs, Information Security, security researchers, hackers, websites, vulnerabilities
1. Introduction
This The National Agency for Protection of Personal Data in cooperation with DARTS Security has
created a plan for increasing the awareness of citizens, a responsibility derived from the Law on
Protection of Personal data, with the purpose of informing the citizens on their constitutionally and
legally guaranteed rights on while data is being processed, as well as their rights to complain whenever
they consider data is not processed in accordance with existing laws, the campaign were called
‘Privacy in Digital Age’ contained a lot of activities related with information security , one of the
activities within the international conference which was held on 27th of January 2015 for the first time
in his format was the workshop ‘The DAY when hacking is legal’. [1]
The purpose of this activity was to raise awareness of Kosovo institutions or companies who deal
with data processing on data security, especially the companies who store or otherwise manage their
data online, including their web sites. In other side stimulating and identifying security researchers
who can legitimately become a white hat hackers. [2]
The supporters of this event were companies who run online businesses. The business profiles various
from web portals, e-commerce, job application, advertisement, etc. The ten numbers of websites has
been tested for security holes by twenty-five hackers. In real-time the links of the websites that have
to be tested were listed in the projector. For each reported bugs to the jury, after the validation the
hacker have been paid per reported bugs.
Juries of three professionals have decided at the end of the activity to give prizes to each hacker, who
has achieved to identify “security bugs” as Critical, Medium or Informative. The jury will not disclose
the contents of the bugs, as it will compromise the security of the companies the web applications of
which have been screened. Hackers have been paid for reported bugs (price per bug). We used to see
similar method so called bug bounty program.[3]
2. Hackers
The Applicants were filtered through interview by Organizer (Darts Security), where interested
hackers applied as individuals but also as a group of three. Each applicant has signed a non-disclosure
agreement that ensures that teams or individuals discovering a security breach or private information
will not disclose the information’s. In NDA was predefined the clausal mentioning the time interval
on which the competitors have the right to scan the websites.
131
From the population of 43 applicants, twenty-five security researchers have been selected among
applied hackers. There have been four groups of hackers within 3 professionals, and the others have
applied as individual.
Questioner containing 11 questions has been answered by 9 hackers attending the workshop.

Six hackers claimed that they were part of some hackers groups in Kosovo and Region.

Enjoyable Academic level with one Master Degree, three of the hackers are still attending
Master, three others have graduated on bachelor degree, the rest were high school students.

The average age of the hackers attending the workshop was 21 years old.

The average age of experience in information security by the hackers was 1.5 years. Starting
from 10 years of information security experience to 6 months.

Anonymously all the hackers liked the format of workshop ‘The DAY when hacking is legal’

By the hackers there is no high cyber security level in Kosovo, even is not enough. Also the
hackers knowledge in Kosovo is ranged as medium.

Four from the nine hackers are actually employed with related information security position.
35
30
25
20
Age
15
Year Experience
10
5
0
A.F I.D B.L G.D R.D F.L A.P A.B A.S
Figure 1. Hackers Average Age compare to year experience in IS
During the workshop there were two issues, one to find and report the vulnerabilities as faster as you
can and the other was the concurrence hackers trying to interrupt each other in their scanning duties.
3. Websites
There were 10 websites selected for Penetration Testing, three of them were from Republic of Kosovo
institutions(public sector) and seven others were from private sector. The idea behind is that the
hackers have always abilities to scan your websites, because the websites are online and could be
accessed everywhere from the internet. But what ‘The DAY when hacking is legal’ did is giving the
websites owners the detailed report not from one hackers logic, but from twenty-five of them. In the
same time the hackers would learn how you can make money even doing the right things like White
Hat hacker.
Table 1. Websites that have been scanned
Domain
Bugs
http://www.rks-gov.net
2
http://www.kryeministri-ks.net
2
132
http://www.valamobile.com/
3
http://www.amdp-rks.org/
NO
http://www.uni-pr.edu
1
http://www.uni-prizren.com
5/6
http://www.tollomed.com/
1
http://www.stikk-ks.org/
2
http://www.pcworld.al/
2
http://www.sollaborate.com
NO
None of the bugs found on the workshop were reported before to websites owners.
4. Vulnerabilities
There were some criteria for hackers to submit the report, they needed to follow the standard required
in order to get the award. The first hacker who reported the bug got the award, duplicate bugs were
not awarded. During eight hours of scanning have been found 19 vulnerabilities, 18 of them were
awarded, 1 was duplicate.
Based on the given time of scanning there are almost three bugs per hour. For more about refer the
table.
10
9
8
7
6
5
4
3
2
1
0
Credential over HTTP
Info Leak
Clickjacking
DoS
XSS
Authentication Bypass
Critical
Medium
Low
Figure 2. Vulnerabilities reported on the workshop
Conclusion
Bug Bounty as the future of Vulnerability Scanning, it gives you the better quality of report. The well
known hackers all over the world not depending geographically can scan and give you a clear report
about vulnerabilities of your product.
Based on parameters like hackers, websites, and vulnerabilities during the workshop we can conclude
that there is medium level of cyber security in Kosovo. In the other side Intelligence Agencies
identified hackers, closely saw how they work, who they are and what can they do.
133
There is a new age of hackers with good experience that can be potentially developed and usefully in
the future, but for the current situation hackers declared that there is no stimulation for the hackers,
there is not enough job positions required for information security.[4]
5.1 Contributors
Thanks go to jury members, that have professionally evaluated the bugs: Prof. Dr. Blerim
Rexha, Mr. Shpend Kurtishaj, Mr. Labeat Avdullahu.
References
1. http://www.pda-ks.com, 20th of October 2015
2. http://www.pda-ks.com/?page=1,8, 21th of October 2015
3. https://en.wikipedia.org/wiki/Bug_bounty_program, 21th of October 2015
4. http://mzhe.rks-gov.net/repository/docs/Lara_Pace.Kosovo-WB.Presentation-June-2015.pdf, 28th
of November 2015
134
Security Concerns of new alternative
telecommunication services
Arbnora Hyseni1, Krenare Pireva1, Miranda Kajtazi2,3
UBT – Higher Education Institution, Prishtina, Kosovo
2Lund University, Lund, Sweden
3Örebro University, Örebro, Sweden
krenare.prieva@ubt-uni.net1, miranda.kajtazi@ics.lu.se2,3
1
Abstract. With the advancing new era of communication, the so-called era of ‘being always online’
many providers offer their services for free. In that sense a small company or a virtual company
becomes a huge competitor for different traditional telecommunication providers. Using the same
services such as: voice calls, video calls, chat and similar services, the internet technology has made
huge changes how users make use of such services . Instead of using these services, users shall install
applications that are specialized for offering these services via applications, such as: Viber,
WhatsApp, Facebook, Google Talk etc.. During the installation and update of these applications
people do not recognize the risks of security and privacy of the information that makes their business
vital, and how such information are abused on the fly and reused for unauthorized purposes. Using
qualitative and quantitative methods we have conducted an empirical study focused on the usage of
“these” services and we have elaborated further the need for increasing the knowledge of people in
the area of security and privacy while using “free” services. This paper will highlight the challenges
by setting a number of privacy and security concerns that are violated while using “free” online
services, also offering a bird’s eye view of numerous recommendations developed by various standard
organizations.
Keywords: Viber, Skype, Facebook, information security, information privacy, online services.
1. Introduction
Telecommunication, as an industry with tremendous changes due to the ongoing evolutions in technology has
experienced new area of communication services, through different open and free service providers. These
providers are offering free apps, through which people could establish voice and video calls, messages,
conference calls etc. The recent generation of service and application companies known as using the “Internet
Over-the-Top services (OTT)” networking as a platform for their service offerings [1]. Companies, such as
Skype, Viber, Facebook messenger, WhatsApp, and many others, have emerged to address the new and
perceived communications needs of both consumer and enterprise users [2].
The mobility of smart devices, and the upgrade of mobile networks to 3G, HSPA+, LTE services and wireless
broadband services, enabled people to be online, most of their time with higher transfer rates. This advancement,
has increased the usage of alternative telecommunication providers, who offer their infrastructure for free. All
of these application use the VoIP standard, for sharing their real-time traffic through Internet. Such new free
communication applications are Viber, Skype, Facebook, Whatsapp etc.
WhatsApp Inc. [3] was founded in 2009 in Santa Clara, California. It is an application which allowes users to
exchange voice, text and multimedia messages through their services. WhatsApp is adaptable and can work
properly in different platforms such as: iPhone, BlackBerry, Android, Windows, and Nokia etc. It uses Internet
for communication
Skype Inc. [3] was founded in 2003, it offers free and commercial services through a single account. It offers
different services, such as chatting, calls, video calls and video conferences, sharing the screen etc. Their
services are also used for education and business purposes, which are using as part of their everyday work. The
application is compatible with different platforms, such as Microsoft, Linux, BlackBerry, iOS, Symbian,
Android etc.
135
Viber [3] was founded in 2010. It is used from more than 606 million people worldwide. It offers services for
calls, video and messaging. It is offered in different platforms, such as iOS, Windows, Android, Symbian etc
Facebook [4] was founded in 2004, in Massachusetts. It launched its new services in Facebook messenger. It
adds video calls, application called Facebook Messenger for chatting and calling as a new services for
communicating peer to peer.
Within this paper, we are trying to identify some of the “hidden” activities within these applications and explore
how the users are using this applications, for what purposes and do they concern for their data security and
privacy.
2. Privacy concerns and identification of new challenges
With the advancement of new era of technology, being able to be online most of the time, there are many
different providers who are offering their services for “free” for telecommunication purposes. In this context,
many small companies, with a small number of employees, are becoming real competitors with huge public
and private telecommunication companies, who are in the same business area. Example, WhatsApp started
with 5 employees its business idea, and now is one of the world competitors in the telecom industry with 700
million users world-wide. On June, 2013, they announced a new record by processing 27 billion messages in
one day [3].
Many users, while using these applications that are being offered as “free”, are scarifying their personal data,
as a paying back the use of their services, where most of the people are not aware of.
Lately, there are raised many privacy concerns while installing these apps, that their data are being used and
reused for unauthorized purposes. While being installed each of the apps required access to users contact list,
so they could discover who is using the app and connect them directly. So in that case, the contact list was
mirrored in their server, including contact information for contacts who are not using these apps [3]. Some of
the apps required even access on messages, galleries etc. In article [5], there were a comparison between
different alternative providers, based on a specific criteria, which is shown in table 1. Such criteria as: Criteria
1: Does the encryption apply while data are transmitted; Criteria 2: Does your communication is encrypted
with any key that the provider could not have access on your data
Criteria 3: Could the contact identities be verified? Criteria 4: Are your past discussion secure, if your
security is broken? Criteria 5: Is the code open to independent review? Criteria 6: Is there a documentation
for security steps? Criteria 7: Has there been a code audit?
Application
Viber
WhatsApp
Skype
Facebook
Messenger
Table 1: Comparison of apps based on 7 Criterias
C_1 C_2 C_3 C_4 C_5 C_6 C_7
Yes
Yes
Yes
Yes
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Yes
No
Yes
So in this context, most of the providers instead of protecting their uses privacy, implemented end-to-end and
transport layer encryption, which are meant to make eavesdropping infeasible even for the service providers
themselves, whereas most of the other criterias’ are neglected [10]. Many security holes where identified last
year’s, Whatsapp faced a problem in the first verion of their apps when the communication was not encrypted,
second security hole was reported which left the users accounts open for session hijacking and packet
analysis[1]. However in viber on July 24, 2013, Viber's support system was defaced by Syrian Electronic
Army, which then was reported from viber that no information was accessed.
3. Methodology
In this research paper we used quantitative method for getting the opinion of the users, analyzing the actual
situation in usage of alternative telecommunication services and how the users are being concern for their data
136
privacy. Data within this research paper are conducted through questionnaires, from four different Universities
and then analyzed and presented further in the result section.
In the questionnaires have participated totally 200 students, from different level of education. Most of the
students are with Computer Science background, which could have an impact on the general results.
4. Results and Discussions
In order to identify how often the users are using the alternative services, and what are the purpose of using
these services from the companies that are offering their application for free, and if they are aware of the
payment back option.
Below are listed some of our graphs and description of our research results done thru couple of questions.
For the question: Do you have knowledge regarding new alternative communication services (Viber, Skype,
WhatsApp, Facebook etc)
Fig 1. Having knowledge of alternative communication services
A 98.8 percent of the users answered with yes, and they already use one of them.
For the question: For whatever purposes do you use them?
Fig 2. Purpose of using alternative communication services
The users declared that 62.4% they used for communicating with friends, followed by 22.4% talking to their
families and relevant.
In the following question, Which of these application do you use more often for the purposed declared on the
previous question?
Fig 3. Using one of alternative communication services
137
From all participated users, 42.4% used Viber, 26.7% WhatsApp, followed by Facebook with 25.5%.
In the questions: (a) How often do you use the alternative services before the traditional options, and (b) Why
do you use them?
(a)
(b)
Fig 4. (a) The use of alternative services before traditional option, and (b) Why using it
For (a) 94 % of users used the alternative services instead of traditional options, whereas for (b) 46.7.5% of
them use because of the low cost, followed by quality and habit.
And in the most important part of our research, we were trying to get the information if the users concern for
their data privacy, for their personal data and also for the traffic that was exchanged. And, 72.9% of them
expressed that they do concern, but it’s a cheaper alternative for generating international and local calls,
followed by 25.9 % of the users that don’t consider at all the privacy issue, as shown in the Fig 5.
Fig 5. The concern of data privacy and security while using free services
And, the final questions: Have you ever doubt on these services for your data privacy and security? And would
you be aware for techniques that will increase the data privacy and security. Surprisingly, in (a) 57.9% of users
declared that they have doubts on which data are having access the providers and how they use, whereas in
(b), 65.9% of the users showed interests in gaining knowledge for increasing the knowledge in awareness
techniques.
(a)
(b)
Fig 6. (a) Doubts about the use of our personal data, and (b) Awareness of using techniques for enhancing
the security of data.
Conclusion
New alternative application for communication are valuable apps and marketing tools. In these applications
are different security risks which could also not be omitted in traditional telecommunication services. However
138
in this direction users are given permission to different authorized and not authorized providers to have access
on their data, violating their privacy and their own security for using their so called “free” services. As
discussed in the last section many users, even that they have doubts on those providers 57.9%, they continually
use their services with 94%, without trying to prevent any security breakage or even to protect their data by
implementing different encryption techniques that could increases the data security.
References and Bibliography
Terry Matthews, “Over-The-Top (OTT) A Dramatic makeover of Global Communications “,
Wesley Clover International, 2014
2. Nuqi, Florian, Pireva Krenare, and Efstathiadis, “The impact of new alternative
telecommunication services on the strategy of traditional telecom providers in Kosovo” , ICCSCE 2014, Durres, Albania
3. Aal, Limbesh B., et al. "Whatsapp, Skype, Wickr, Viber, Twitter and Blog are Ready to
Asymptote Globally from All Corners during Communications in Latest Fast Life." Research
Journal of Science and Technology 6.2 (2014): 101-116.
4. Jones, Harvey, and José Hiram Soltren. "Facebook: Threats to privacy."Project MAC: MIT
Project on Mathematics and Computing 1 (2005): 1-76.
5. Mazurczyk, Wojciech, and Zbigniew Kotulski. "Lightweight security mechanism for PSTNVoIP cooperation." arXiv preprint cs/0612054 (2006).
6. Mahajan, Aditya, M. S. Dahiya, and H. P. Sanghvi. "Forensic analysis of instant messenger
applications on android devices." arXiv preprint arXiv:1304.4915 (2013).
7. Schrittwieser, Sebastian, et al. "Guess Who's Texting You? Evaluating the Security of
Smartphone Messaging Applications." NDSS. 2012.
8. Coull, Scott E., and Kevin P. Dyer. "Traffic Analysis of Encrypted Messaging Services: Apple
iMessage and Beyond." ACM SIGCOMM Computer Communication Review 44.5 (2014): 511.
9. Coull, Scott E., and Kevin P. Dyer. "Traffic Analysis of Encrypted Messaging Services: Apple
iMessage and Beyond." ACM SIGCOMM Computer Communication Review 44.5 (2014): 511.
10. Ahmar Ghaffar, “How Secure Is VoIP?”, [Online in: http://www.isoc.org/ pubpolpillar/voippaper.shtml, Accessed on: 12 January 2016]
11. TBU News. (2015). Privacy and Security – Face Viber And BBM not quite Secure Tor And
Open Whisper Champions of Security [Accessed on, 12 January 2016]
1.
12. Kierkegaard, Sylvia (2006). "Blogs, lies and the doocing: The next hotbed of
litigation?". Computer Law and Security Report 22 (2): 127.
139
Implementation of the AODV Routing in an Energyconstrained Mesh Network
Altin Zejnullahu1, Zhilbert Tafa2
1
Dep. of Software Engineering and Telecommunications University of Sheffield,
Thessaloniki, Greece
2
Dep. of Computer Science and Engineering UBT – Higher Education Institution,
Prishtina, Kosovo
zhilbert.tafa@ubt-uni.net2
Abstract. Wireless sensor networks (WSNs) compose the fundamental platform for a number of
Internet of Things (IoT) applications, especially those related to the environmental, health, and
military surveillance. While being autonomous in power supply, the main challenge in node’s
processing and communication architecture design remains the energy efficiency. However, this goal
should not limit the main functionality of the system which is often related to the network coverage
and connectivity.
This paper shows the implementation of the Ad-hoc On-demand Distance Vector (AODV) routing
algorithm in an XBee based platform. As shown, the network can achieve low power consumption
per node primarily due to the energy efficiency of the wireless transceivers and the due to the
capability of the firmware to enable different operation modes. On the other hand, while inheriting
the advantages of flooding-based route discovery protocols, the implemented AODV algorithm
further minimizes the data and processing overhead, which implies the additional lifetime
prolongation of the energy-constrained mesh network.
Keywords: AODV, routing, wireless sensor networks, XBee
1. Introduction
Wireless Sensor Networks and Mobile Ad Hoc Networks (MANETs) have attracted great academic
and engineering attention due to the technology trend convergence towards the concept of IoT. A
Wireless Sensor Network indicates a particular type of network which is characterized by a distributed
architecture created from a set of autonomous electronic device which are able to sense, scan and pick
up information from their surroundings and able to communicate with each other [1]. Recent
technological advances have enabled wireless communications and digital electronics to be able to
develop small low-power devices, which are multifunctional and able to communicate with each other
via wireless technologies [2]. A typical WSN represents a wireless network composed of small powerautonomous nodes that sense the physical parameters from the environment, locally process them,
and sends the sensed data wirelessly towards the destination. In addition to the constraints implied
from the limited ICT resources, WSN’s application requirements include self-organizing capabilities
as well as robustness to the electromechanical influences and to the topology changes. With the same
fundamental constraints, MANETs additionally presume the higher level of mobility, which again
implies slightly different approaches in network design. With the given application requirements, the
MANETs networking protocol design goes far beyond the well-established networking protocols that
are present in today’s Internet. Since nodes are usually battery supplied (or alternatively by some
other means of limited sources), the main goal in the design of the protocols is minimizing the power
consumption in a relatively highly dynamic networking topology, while optimizing the other ordinary
application requirements such as network coverage, link capacity, latency, network robustness,
reliability, security, etc. This approach is used in most of the protocol design, in each of the
networking layer.
140
Routing in MANETs is a challenge that, due to the mentioned limitations, is still not standardized and
is highly application-specific. There are many routing protocols proposed in literature, but only few
of them are implemented. The Ad-Hoc On- Demand (AODV) routing protocol is one of the most
popular reactive protocol for ad hoc networks.
This paper presents an implementation and energy-focused analyze of the AODV protocol in an XBee
enabled platform.
The paper is structured as follows. Section 2 presents the main WSNs’ challenges and routing
philosophies. Section 3 presents the experimental setup while the appropriate results are briefly
discussed in Section 4. Finally, Section 5 concludes the paper.
2. Mesh networks and routing
Mesh networks offer scalable and flexible architecture, since the number of deployed sensor nodes
may vary from couple to hundred, thousands or millions. Accordingly, the design of the
communication protocols must be focused on enabling new concepts such as ad-hoc clustering, multihop routing, energy efficiency, etc. Furthermore, the network must preserve its stability even when
new nodes are introduced into the network, which means that additional communication messages
will be exchanged in order for the new nodes to be integrated into the network [3].
2.1 Main challenges in WSNs
The requirements of each WSN or MANET based application may vary significantly. In most cases,
several elements should be considered in order to make the best design decisions.
Fault tolerance is one of the main attributes of the network, especially important in applications where
the network is deployed in the inaccessible regions such as forests, battlefields, etc. It presents the
network capability to function without interruption even in case of failure of one or more sensor nodes.
In this situation, adaptive network protocols should be implemented. For example, the AODV
protocol is able to establish new links in case of node failures or in case a better route is available [4].
Network deployment can also be categorized in two ways: nodes can be densely or sparsely deployed
[4][2]. Also, sensor network can be deployed randomly or in deterministic manner. The critical WSN
implementation challenges raise when the network is dense and when the nodes are deployed
randomly. In these cases the protocols should be ready to deal with the issues such as coverage,
connectivity, collisions, interference, traffic overhead (in the process of creating and maintaining the
routes), energy efficiency, etc. The AODV protocol, by its functionality nature, enables for the
addressing of most of last issues.
Power consumption remains one of the main challenges in WSNs and MANETs. The main issue is
implied from the fact that a wireless sensor node is a microelectronic device equipped with a limited
amount of power source [5]. These devices are expected to be of miniature physical dimensions, to
meet the ICT requirements and, at the same time, to live for as long as possible while supplied from
the battery or from the alternative energy sources. Hence power conservation and power management
are the important issues of the architecture design. Due to this reason, researches are mostly focused
on the design of power aware protocols and algorithm for sensors networks [5]. The Digimesh
firmware that is used in the experimental setup in the proceeding section, and which relies on AODV
philosophy, enables for the execution of power aware multi-hop routing.
Besides the given challenges and constraints, WSNs pose many additional limitations (that are
usually considered as secondary) such as the implementation cost, latency, complexity, etc.
2.2 Routing and AODV
Routing techniques are required in order to enable the data path between sensor nodes and the sink
node/s [6][7]. Based on various parameters including their mode of functioning and type of target
applications, the protocols can be classified as proactive, reactive and hybrid [6].
141
2.2.1 Routing protocols and mesh networks
In the environments where a proactive protocol is used, the communication between nodes and/or
gateway is performed through a predefined route. Hence, known as table-driven routing protocol, a
proactive protocol serves as a requirement establishment for nodes which need to be able to equip
their routing table with the latest data in order to be able to establish and perform any required
connection [8]. So, whenever a node needs to send a specific packet, there is no need to search for the
route. Instead, the route is already there. This implies for the high delay reduction. However, there
are many disadvantages of this routing philosophy. First, the protocol efficiency is poor if the topology
changes rapidly [9]. Second, the routes are actively and periodically maintained, and the entries in the
routing tables exist even though the routes might not be active for a while. Furthermore, these routing
protocols constantly maintain routes (which even may never be used) and consume bandwidth to keep
routes up-to-date for bringing the only advantages – provision of little or no delay for route
determination [6]. Some representative types of proactive routing protocols are [7]: Optimized Link
State Routing Protocol (OLSR), Destination Sequenced Distance Vector (DSDV), Wireless Routing
Protocol (WSR).
On the other hand, in order to discover a route, the reactive routing protocols flood the network. This
decreases the bandwidth utilization but scales well in frequently changing network topologies [9].
However, the route maintaining process does not incur much overhead, since the route maintenance
is limited to only those destinations which are target of receiving data. This construction leads to a
conclusion that reactive routing protocols are best suited for networks with higher nodes’ mobility.
Some representative reactive or On-demand routing protocols [6] are: DSR (Dynamic Source
Routing), TORA (Temporally Ordered Routing Algorithm), ABR (Associativity Based Routing),
AODV (Ad-Hoc On-Demand Distance Vector).
Typically, the networks which operate under hybrid routing protocols are separated into specific
regions and/or zones [10]. Here, the routing protocols are created as a result of mixing the usage of
proactive and reactive routing protocols [8][6]. Each node maintains routing information for a specific
zone. That is, communication between nodes is kept updated in a proactive way. On the other hand,
zones are interconnected with each other by using reactive routing protocols which is highly
bandwidth efficient, especially in case of rarely changing inter-zone communication. Hybrid
approaches provide a compromise on scalability issue in relation to the frequency of end to-end
connection, the total number of nodes, and the frequency of topology change.
2.2.2 The AODV
The AODV protocol reduces control traffic by originating path requests only on demand. To monitor
active neighbors, AODV uses stateless hello messages. Upon failure, it generates an error message to
notify upstream sensors that use the broken path to avoid it [11]. When a broken link is discovered
and an alarm is triggered to notify that a prior valid path now it is not usable some processes are
triggered in order to retain the path to the destination. Even though the path monitoring and route
table management are as important as the path construction, this section will cover only route creation
process. An event or alarm will trigger a route request only if a valid path to the sink doesn’t exist[12].
Paths expire either implicitly, e.g. routing state timeouts, or implicitly, e.g. error packets sent. Sensors
buffer their data packets until they receive a reply [13]. There can be more than one outstanding
request at a time, identified by a sequence number. Route requests are broadcast [14], constrained by
a TTL hop count limit which is used to reduce flooding redundancy [15]. Maximum path length is its
default value [14]. All intermediate relays compete with each other to become part of the best path;
they increment an additive metric (e.g. hop count) and rebroadcast the request; meanwhile they cache
the sender as the next hop to the originator. The AODV protocol uses implicit acknowledgements to
determine bidirectionality between any pair of sensors [11]. If a sensor marks, or blacklists, a link as
unsteady, it ignores it in the path discovery process [1][12].
Once the request is receipt, a path is established by sinks in reverse: thus the reply is propagated by
them using as the next relay to the originator the previous sensor [30][7]. A sink is capable of receiving
more than one request per source, even though the first request is able to traverse the path fastest [6].
Moreover, the gateway or sink is able to reinforce multiple paths toward the originator. Depending
on the strategy which is being used, these paths can be discarded, used, or stored. Reply messages are
unicast, rather than multicast [12][11]. Sensors are able to generate free replies, in case they already
142
maintain an active path toward the destination [6]. This does not guarantee optimality, but improves
responsiveness [6]. Nevertheless, it is important to notify sinks for path interests in order to monitor
abnormal traffic patterns [6][9].
The AODV protocol offers a good balance between minimal overhead, good network reaction on
dynamic route changes and acceptable latency. The mentioned reasons make it appropriate for the
majority of MANET-based applications.
3. Experimental setup
The experimental scenario includes two sensor nodes and a sink node, as shown in fig. 1. Wireless
communications are achieved by using XBee standard/modules while the mesh topology construction
and the AODV implementation are provided by Waspmote Digimesh firmware [17].
Fig. 1: Network topology
Digimesh incorporates different modes of operations such as low power sleep modes with
synchronized wake up, low power mode with variable sleep time, and low power mode with variable
wake up time. This provides for the high power efficiency. It is also self-healing, i.e., the nodes can
be added and removed without consequences in the sense of network connectivity. Finally, AODV is
designed for ad-hoc communication manner and enables for the reliable data delivery by using
acknowledging approach.
The aims of the experiments were to implement AODV/Digimesh protocol in performing multi-hop
routing for energy conversation and communication coverage improvements. The case study is
limited on temperature sensing.
So, the experiment has three phases:
a) Temperature measurement in one node.
b) The AODV execution, i.e., the executing of the AODV routing from end to end and sending
the temperature measurements periodically.
c) Measuring the energy efficiency.
4. The results
The energy consumption was measured per node. Measurements were performed in both cases - when
nodes are sending data in static point-to-point manner, and when the data dissemination is made in
multi-hop manner with AODV implemented.
143
For measuring the power consumption, the ampermeter is connected in series to measure the current.
In both scenarios, the current is measured using same conditions, including nodes’ distance, packet
size etc. The current is measured in both active and sleep mode and the results are showed in table 1.
Table 1: Current drain with and without AODV implemented.
Active Mode
49 mA
One Hop routing
54 mA
Multi-hop routing using
AODV routing protocol
Sleep Mode
0.7 mA
0.9 mA
As can be noted, while being very efficient in dealing with the communication issues in dynamic
mesh environments, the AODV protocol is also very energy efficient, which is one of the most
important issues in WSNs and MANETs.
Conclusions
The paper presents an implementation of AODV routing protocol in temperature measurement. The
analysis was focused on power efficiency of the AODV in multi-hop mesh networks. In order to
derive the “energy weight” of the AODV, we have compared the power consumption of AODV
enabled network with the one that does not use the AODV, using predefined same testing conditions
and scenarios. From the derived measurement, it can be easily concluded that, besides its advantages
in bandwidth utilization, relatively low latency, and a good response on topology changes, AODV
protocol is also a “light” one in the sense of energy consumption. Hence, it is found to be suitable for
use even in those applications where the power conservation is one of the primary limitations.
References
1
ChiaChiarara Buratti, Andrea Conti, Davide Dardari, Roberto Verdone, " An overvi overview
on Wireless Sensor Networks Technology (ISSN 1424-8220;CODEN:SENSC9)," 2009
2
R. Flickenger, "Wireless Networking in the Developing World," 2006.
3
Dunfan Ye, Daoli Gong, Wei Wang, "Application of Wireless Sensor Networks in
Environmental Monitoring," IEEE, 2nd Internation Conference on Power Electronics and
Intelligent Transportation System, 2009.
4
Bruno R., Conti M., Gregori E., "Mesh Networks: Comodity multihop Adhoc Networks,"
IEEE, Communication Magazine , vol. 43, 2005.
5
Chulsung Park, Kanishka Lahiri, "Battery discharge characteristics of Wireless Sensor Nodes.
An experimental analysis," Proceedings of the IEEE Conference on Sensor and Ad-hoc
communications and Networks (SECON), 2005.
6
Alexandros Koliousis, Joseph Sventek, "Proactive vs reactive routing for wireless sensor
networks".
7
Abdul Hadi Abd Rahman, Zuriati Ahmad Zukarnain, "Performance Comparison of AODV,
DSDV and I-DSDV routing protocols in MObile Ad Hoc Networks," European Journal of
Scientific Research ISSN 1450-216x, , vol. 31, pp. 556-576, 2009.
8
Hmalik M., Awais Q, Jamil M., Dhyani A., "Performance analysis of proactive and reactive
protocols in mobile Ad-hoc networking," Islamabad 22-24, April, 2014.
9
Anjali Bhatt, Prashant Bhati, "Performance Analysis of reactive and proactive routing
protocols for Vehicular Ad-hoc Networks," International Journal of Electrical and Electronics
Research, vol. 2, no. 4, 2014.
10 Jamal N., Al-Karaki, Ahmed E. Kamal, "Routing Techniques in Wireless Sensor Networks".
11 Charles E. Perkins, Elizabeth M. Royer, "Ad Hoc on Demand Distance Vector Routing".
144
12 Ian D. Chakeres, Elizabeth M. Belding-Royer, "AODV Routing Protocol Implementation
Design".
13 Elizabeth M. Royer, Charles E. Perkins, "An implementation study of the AODV routing
protocol".
14 Iftikhar Ahmad, Uzma Ashraf, Sadia Anum, Hira Tahir, "Enhanced AODV route Discovery
and Route Establishment for QOS provision for real time transmission in MANET,"
Interantional Journal of COmputer Networks & Communications , vol. 6, 2014.
15 Venetis Kanakaris, David Ndzi, Kyriakos Ovalladis, "Improving AODV Performance using
dynamic density driven route request forwarding".
16 Guoyou He, "Destination-Sequenced Distance Vector (DSDV) Protocol".
17 Libelium, "Waspmote Digimesh Networking guide".
145
Exploring the role of sentiments in identification of active
and influential bloggers
Mohammad Alghobiri1, Umer Ishfaq2, Hikmat Ullah Khan3,* , Tahir Afzal Malik4
1,2
Department of Management Information Systems, Ibn Rushd College Management
Sciences, Abha, Kingdom of Saudi Arabia
3,4
Department of Computer Science, COMSATS Institute of Information Technology,
Attock, Pakistan
maalghobiri@kku.edu.sa1, tahir.malik@ibnrushd.edu.sa2, umer.bravo@gmail.com3,
hikmatullah@comsats.edu.pk4
Abstract. The social Web provides opportunities for the public to have social interactions and online
discussions. A large number of online users using the social web sites create a high volume of data.
This leads to the emergence of Big Data, which focuses on computational analysis of data to reveal
patterns, and associations relating to human interactions. Such analyses have vast applications in
various fields such as understanding human behaviors, studying culture influence, and promoting
online marketing. The blogs are one of the social web channels that offer a way to discuss various
topics. Finding the top bloggers has been a major research problem in the research domain of the
social web and big data. Various models and metrics have been proposed to find important blog users
in the blogosphere community. In this work, first find the sentiment of blog posts, then we find the
active and influential bloggers. Then, we compute various measures to explore the correlation
between the sentiment and active as well as bloggers who have impact on other bloggers in online
communities. Data computed using the real world blog data reveal that the sentiment is an important
factor and should be considered as a feature for finding top bloggers. Sentiment analysis helps to
understand how it affects human behaviors.
Keywords: Blogger, Sentiment, Social web, Big Data.
1. Introduction
The emergence of the social web has revolutionized the ways people communicate and share views
with each other. It offers a number of well-known services which include blogs, media sharing
networks, wikis, etc. These sites allow people to connect with each other and share information. In
this way, users of social media can access real time, valuable information [1]. Users get the advice of
others before making decisions for example, choosing a place for shopping or buying products of a
particular brand. They listen and trust their opinions. In this way, they can be influenced by others
whom they are connected with and such users are termed as influential bloggers. Identification of
such users has been a real challenge [2].
People share comments on the blog posts and these comments are displayed in reverse chronological
order. Blogosphere is a broad term which covers all the blogs on the web. Blogs, usually have two
categorized. Community blogs allow people to publish their blogposts or comments or both whereas
individual blogs are like personal diaries. Blogging has become very popular and people use to it
for voicing their opinions, reporting news or mobilizing political campaigns [3][4]. The bloggers
usually create their interest groups and play a unique and significant role in social communities. The
democratic nature of blogosphere has made it a lucrative platform for social activities and influence
propagation [5]. The identification of influential bloggers has direct applications in online marketing
and e-commerce. By gaining the trust of influential bloggers, companies can turn them into their allies
and can save huge advertising sum [6]. Similarly, the identification of influentials can help in learning
the market trends and improving insights for better custom care. Influential can further assist in
reshaping the businesses and rebranding their products [7]. In this work,we find the role of sentiments
146
and check how the sentiment of blog content correlate with the activity and recognition of bloggers.
In other works,we find active as well as influential bloggers and then check the sentiment of the
bloggers and find the top bloggers with respect to activity, influence within blogging community and
check the role of sentiments too. The checking of sentiment in this research domain is a novel
contribution. In the following sections, first we present the relevant work in section 2. We then outline
the proposed framework where the research methodology is discussed in detail in section 3. The
evaluation results are discussed in section 4. Finally, we conclude with section 5 and also give
direction for our future work.
2. Related Work
As blogging has gained considerable momentum in todays world and people are influenced by the
opinions of their fellow bloggers. Researchers have highlighted various issues arising due to the
plathore of information piling up on the web caused by social media revolution and emphasized the
need of developing intelligent tools for mining valuable information [1]. Since, social media contains
huge amounts of information about people, their customs and traditions. This valuable information
can be exploited for better understanding individuals and their communities [8]. Studies show that
influence has multiple dimensions and types. To differentiate one type from another and evaluate its
importance over the other is real challenge [9]. PageRank is a well-known algorithm for ranking pages
on the web. Zhou At el. [10] modified PageRank by taking into account nodes on the social media
network for identifying the opinion leaders. The authors in [11] identified the influentials by ranked
the blogs. Various blogs on the blogosphere are compared and their relative importance is
quantitatively measured. The work proposed in [12] developed a data mining based model for
identifying the influential bloggers. It calculated the potential influence of influentials to maximize
the spread of information on social media. Most of the people on social media act as information
consumers. Research presented in [2] proposed a model which identifies the influential bloggers on
the basis of information they produce, rather than their popularity within the social community. A
detailed discussion is present in [3] which critically analyzed well-known blogger ranking algorithms,
discussed their shortcomings and introduced a new ranking metric. We find a standard model [14]
which is based on four basic charateristics of a blogger. It is novel in this regard that it finds influential
bloggers and compared with active and concludes that both are important.Then it compares with
pagerank as well and finds the model is better instead of pagerank to find top bloggers. This work is
continuation by the same authors who actually initiated this domain of finding top influential bloggers
by introducing their model known as iIndex [16]. Then, another two metrics have been proposed in
another research work [15] which considers time as an important factors and measures the temporal
order of the bloggers in a better manner.
3. Proposed Approach
The model proposed by Agarwal et al. [14] is considered as a standard model. In our work, we are
extending this model by adding a novel feature i.e. sentiment score. We are proposing a model based
on the features of activity, recognition and sentiment score.
Activity of bloggers is measured by examining their number of blog posts. It is an effective method
which represents the productivity of bloggers [5]. Active bloggers, usually have a large number of
blog posts. Whereas the recognition represents the position or status in the social community. The
number of inlinks (citations) and comments on a blog post is taken as the recognition of the blogger.
Large number of comments on a blog post show the interest of other bloggers as they care to write
comments [14]. Thus, recognition can be a good feature to measure the influence of a post on other.
It is recognized by many researchers due to its simplicity and efficiency [5] [15] [14] [16].
Though the above-mentioned features like the number of blog posts, inlinks and comments can
effectively identify influential bloggers, but adding a new feature i.e. sentiment score, can increase
the accuracy of the model as discussed in section. 4. We are exploring the role of sentiments in
identification of active and influential bloggers and examining how the sentiment score of a blogger
147
is correlated with his/her activity and recognition. To the best of our knowledge this feature is not
considered yet. Table. 1 shows the features included in the proposed model.
Table 3. Features of the proposed Model
Sr
Feature Name
1. Number
1
of blog posts
2. Number of inlinks to a blogpost
3. Number of comments on a blog post
4. Sentiment Score of a blogger
4. Results Discussion
In this section, we examine how the sentiment score correlates with activity (number of posts) and
recognition (inlinks and comments) of a blogger. We performed this experiment on Engadget11, a
technology blog with real time features of a blog site. The characteristics of the dataset are shown in
table 2.
Table 2. Dataset Characyeristics
Engadget
Bloggers
93
Posts
63,358
Inlinks
319,880
Comments
3,672,819
Figure. 4 Snapshot of SentiStrength v2.2
Table 3 below presents the results of our findings. The last three columns show the sentiment score,
activity and recognition score of the top-10 bloggers of Engadget dataset. The blog posts of a blogger
represents his/her activity. Inlinks and comments on a post together constitute the recognition of a
blogger. Sentiment score, on the other hand, is calculated using SentiStrength. Figure. 1 below shows
a snapshot of SentiStrength. In the model proposed, we calculated the sentiment score of blog posts
using a mining based scaling system. A document or a sentence usually has words with positive,
negative, or neutral sentiment. These words are analyzed on a scaling system and assigned positive
11
http://www.engadget.com
148
and negative values. The system assigns values from a dictionary which interprets the positivity or
negativity with human level accuracy [17].
Table 3. Bloggers ranking based on activity
1
No.
of
blogposts
Darren
Murph
Inlinks
Comments
Activity
Recognitio
n
Sentiment
Darren
Murph
Thomas
Ricker
Darren
Murph
Darren
Murph
Laura June
Darren
Murph
Ryan Block
Ryan Block
2
Peter Rojas
3
Ryan Block
Paul Miller
Laura June
Paul Miller
4
Paul Miller
Nilay Patel
Joshua
Topolsky
Peter Rojas
Ryan Block
Paul Miller
5
6
Donald
Melanson
Thomas
Ricker
7
Nilay Patel
8
Evan Blass
9
1
0
Joshua
Topolsky
Chris
Ziegler
Donald
Melanson
Joshua
Topolsky
Chris
Ziegler
Nilay Patel
Thomas
Ricker
Donald
Melanson
Donald
Melanson
Thomas
Ricker
Darren
Murph
Thomas
Ricker
Paul Miller
Ryan Block
Paul Miller
Peter Rojas
Joshua
Topolsky
Donald
Melanson
Thomas
Ricker
Nilay Patel
Ryan Block
Nilay Patel
Joshua
Topolsky
Donald
Melanson
Chris
Ziegler
Ross Miller
Chris Ziegler
Chris Ziegler
Vladislav
Savov
Peter Rojas
Evan Blass
Nilay Patel
Ross Miller
Evan Blass
Joshua
Topolsky
Chris
Ziegler
In order to check the correlation of sentiment score with the modules of activity and recognition, we
are using the famous Pearson correlation, one of the highly used performance evaluation measure.
The Pearson correlation indicates the linear correlation between the two quantities being measured.
Table 4. Spearman's Correlation of the three modules
Spearman Correlation
Kendall Correlation
Activity vs. Recognition
0.4345
0.12
Activity vs. Sentiment
0.9936
0.91
Recognition vs. Sentiment
0.4853
0.2
The above table shows that activity and sentiment have strong relationship. This means that changes
in activity are strongly correlated with changes in sentiments. Though, the other two i.e. activity vs.
recognition and recognition vs. sentiment show a positive correlation. But their correlation values
show that they don’t have a very strong relationship.
Conclusion
In this paper, in addition to activity and influence, we also focus on the sentiment analysis of blog
posts. We use the blog post content, compute their sentiment scores using standard methods and
correlate them with the characteristics of the top bloggers who are active as well as influential. Several
existing and proposed features have been used. The results computed using the real world blog data
“Engadget” reveal that the sentiment is an important factor and should be considered for identification
of influential bloggers in the blogging community. For future work we aim to combine our proposed
solution into a framework for extending the modules of activity, recognition and sentiment score by
149
adding more features to create a comprehensive model which enhances the task of identifying the
active and influential bloggers.
150
References
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
N. Agarwal and H. Liu, "Blogosphere: research issues, tools, and applications," ACM
SIGKDD Explorations Newsletter, vol. 10, no. 1, 2008.
D. M. Romero, W. Galuba, S. Asur and B. A. Huberman, "Influence and passivity in social
media," in 20th international conference companion on World wide web, Hyderabad, 2011.
J. Bross, K. Richly, M. Kohnen and C. Meinel, "Identifying the top-dogs of the blogosphere,"
"Standpoint in Political Blogs: Voice, Authority, and Issues," Women's Studies: An interdisciplinary journal, vol. 40, no. 3, pp. 269-298, 2011.
N. Agarwal, D. Mahata and H. Liu, "Time- and Event-Driven Modeling of Blogger Influence,"
in Encyclopedia of Social Network Analysis and Mining (ESNAM), New York, Springer,
2014, pp. 2154-2165.
B. Sun and V. T. Ng, "Identifying influential users by their postings in social networks," in 3rd
international workshop on Modeling social media, Milwaukee, 2012.
G. Mishne and M. d. Rijke, "Deriving wishlists from blogs show us your blog, and we’ll tell
you what books to buy," in 15h International conference on World Wide Web, Edinburgh,
2006.
S. Kumar , N. Agarwal , M. Lim and H. Liu, "Mapping socio-cultural dynamics in Indonesian
blogosphere," in Third International Conference on Computational Cultural Dynamics,
Washington DC, 2009.
J. Tang, J. Sun, C. Wang and Z. Yang, "Social influence analysis in large-scale networks," in
15th ACM SIGKDD international conference on Knowledge discovery and data mining, Paris,
2009.
H. Zhou, D. Zeng and C. Zhang, "Finding leaders from opinion networks," in IEEE
International Conference on Intelligence and Security Informatics (ISI), Dallas, 2009.
X. Song , Y. Chi, K. Hino and B. Tseng, "Identifying opinion leaders in the blogosphere," in
6th ACM conference on Conference on information and knowledge management, Lisbon,
2007.
Y. Singer, "How to win friends and influence people, truthfully: influence maximization
mechanisms for social networks," in Fifth ACM international conference on Web search and
data mining (WSDM), Seattle, 2012.
T. A. Malik, H. U. Khan and S. Sadiq, "Dynamic Time Table Generation Conforming
Constraints a Novel Approach," in International Conference on Computing and Information
Technology (ICCIT), Al-Madinah Al-Munawwarah, 2012.
N. Agarwal, H. Liu, L. Tang and P. S. Yu, "Modeling Blogger Influence in a community,"
L. Akritidis, D. Katsaros and P. Bozanis, "Identifying the Productive and Influential Bloggers
in a Community," IEEETransaction on System, Man, Cybernetics, Part C, vol. 41, no. 5, pp.
759 - 764, 2011.
N. Agarwal, H. Liu, L. Tang and P. Yu, "Identifying the influential bloggers in a community,"
in in: Proceedings of the International Conference on Web Search and Web Data Mining, New
York, 2008.
M. Thelwall, K. Buckley, G. Paltoglou, D. Cai and A. Kappas, "Sentiment strength detection in
short informal text," Journal of the American Society for Information Science and Technology,
vol. 61 , no. 12, p. 2544–2558, 2010.
18
151
Internet of Things: From applications, challenges and
standardization to Industry implementations
Xhafer Krasniqi
UBT-Higher Education Institution
NEC Corporation
xhafer.krasniqi@emea.nec.com
Abstract. The Internet of Things that is defined as anything that can be accessible anytime and
anywhere provides connectivity to different objects and sensors around us and which will enable the
transfer of different data between these objects and devices. A thing in the Internet of Things can be
any natural or man-made object that can be assigned an IP address with a capability to exchange date
over a network. There is a huge number of applications of IoT to benefit users, such as health
monitors, smart homes, connected cars etc.
If everything around us is connected and information about these things that can contain sensitive
information, e.g. health and other personal information, are collected then these networks become
very important and must be able to provide a proper security and privacy. It is believed that by 2020
there will be over 50 billion things that could be connected to Internet.
Internet of things are very much associated with M2M (machine to machine communication) that is
identified as a technology that makes objects smart, like smart homes, smart utility meters etc. M2M
actually is considered to be a subset of IoT and which is mainly used for difficult and dangerous tasks,
e.g. nuclear plants, etc. The deployment of IoT has already started and is expected to transform the
way we live. According to Gartner, a technology research company, the Internet of Things has just
reached the deployment stage by early adopters and the full deployment is expected in over ten years.
From an industry angle, this paper will examine the market and technical trends of Internet of Things,
main applications that will be supported by this technology, key issues and challenges faced by the
industry, standards activities around IoT and finally the implementation landscape.
Keywords: Internet of Things, M2M, nuclear plants, tasks.
1. Introduction
The idea of connecting everything existed for quite some time, since early 2000, but the technology
to support this connectivity and other supporting factors were not available then. The idea came from
the need to have computers manage all things in daily life, those inanimate and living things, that
could be equipped with identifiers and wireless connectivity, and which could communicate with each
other. Although we do not have a clear definition of IoT yet, probably the best definition of IoT that
the industry has come up with is that the IoT is the disruptive convergence of technology,
telecommunications, data, and people as reflected in the figure below. So, IoT is a system of enabling
technologies and not a specific set of technologies.
152
Fig. 1. Internet of Things (Source: datascience.com)
Back in 2000 we did not have enough addresses for all those devices, we did not have good mobile
data coverage, mobile service was extremely expensive and the battery technology was not able to
power up and support all these objects. The situation with those factors has changed drastically in the
meantime and the prices have fallen rapidly as shown in figure 2.
Fig. 2. Relevance of price falling
Resulting from these condition changes, the industry is thinking and getting seriously involved in
developing and implementing the IoT. It is believed that by 2020 there will be over 50 billion things
that could be connected to Internet. To speed up and achieve the full deployment the industry still
needs to see the innovative IoT solutions from vendors and others involved which would enhance the
customer experience and improve the operating efficiency. We also need to develop common
standards to enable communication and interaction between different products and different solutions
from different vendors and players and above all we also need close collaboration between the
industry stakeholders to achieve this full deployment.
This paper focuses on the vertical markets of IoT, applications and drivers of this technology, then
technical and market trends, challenges and issues faced by the industry, standards around IoT and
finally use cases and implementation landscape.
153
2. IoT Vertical Markets
Given that IoT applies to a wide range of markets, it is important to identify those markets
in order to better categorise the applications and to better understand their functions.
Main vertical markets where IoT technology and IoT applications are having a huge impact
are the following:






Computing industry
o Human behaviour towards internet connected devices can change by installing
sensors in computing devices to mimic our senses
Health industry
o Nano-sensors will be included in the medicine that the patient is taking and
doctors can monitor the effect of it on the patient in real-time
o It is anticipated that by 2020 patients will not have to go to a doctor for a doctor
to know how the patients are doing, as the doctor receive all information from
the sensors the patient is wearing [2]
Environment
o Sensors will be placed around the work to improve the response time to
earthquakes
o To help find more oil in different corners of the world
Household
o IoT enhances home automation system to integrate electrical devices in a house
with each other and gives the users the possibility to access those devices via
smartphones or tablets
Automotive
o Sensors have found and are finding wide application in transportation industry
to monitor driver behaviours, to optimize routing and fuel consumption
Retail market
o This is another vertical market where RFID tags are being installed widely to
improve the efficiency of checking out of items which do not have to be put on
the conveyer belt to be scanned, but they can be scanned automatically by
leaving the supermarket
3. Main Applications And Drivers Of IoT
The list of applications where IoT can be applied and used is huge since trillions of sensors will
comprise the IoT. This will drive to 50 zettabytes of data annually by 2020 and which will be an
enormous task to collect and analyse [2].
For today’s Internet humans are the main creators of data, e.g. people write documents, build
databases, photograph and upload images, but by 2020 we will see almost the exact opposite, as
hundreds of millions of Internet-connected devices primarily use information that originates from
other devices and not from humans. This is also known as machine-to-machine (M2M)
communications, a subset of IoT and fast-growing portion of the embedded systems market.
The list of applications and drivers can be categorized in few groups based on the use and the markets
that push for full deployment. This list can serve as a subgroup of IoT vertical markets, but it is more
granular and function-specific:

Tracking for the purpose of advertising and marketing
o Helps advertisers and marketers to make creative campaigns by tracking
consumers and their behaviour

Enhanced context awareness
o Enables a capability to sense their physical environment, and adapt their
behaviour accordingly

Process and business optimization
o IoT is widely accepted as a technology that would help to optimize processes
and businesses
154






Business transformation
o IoT will transform businesses and improve their operation cost
Optimization of resource consumption
o Enhances resource consumption, e.g. fuel consumption in the automotive
systems
Enhanced control in autonomous systems
o Improvement of automation process through data analysis collected by IoT
sensors
Ubiquitous networking
o Allows program applications to follow a user through IoT special sensors
wherever he/she would go
Connected computing
o Our behaviour recording through all our connected devices
Ubiquitous sensors
o Low sensor prices is considered to be an important driver
o Prices will move from $0.09 to $0.03 for passive sensors and from $1 to $0.15
for active sensors respectively by 2021[2]
o These low prices of sensors can lead to widespread deployment of sensors in all
types of devices and all types of industries
Fig. 3. Individual technologies vs. IoT (Source: Forrester Consulting, Nov. 2014)
4. IoT Market And Technology Trends
As predicted there is a huge potential for IoT and the industry’s expectation is shown in figure 4,
where IoT is one of the biggest technology waves since 1960s. Internet of things is considered to be
the third wave of Internet after the first wave that brought common software applications like
Microsoft Windows and Google and second wave that enabled transition to mobile software such as
iOS and Android that enabled large number of devices to easily communicate.
155
Fig. 4. IoT as a big technology wave (Source: KPCB, ITU, MS Research)
According to Gartner, a technology research company, the Internet of Things has just reached the
deployment stage by early adopters as it was the case with Cloud computing in 2008-2010 and with
Big Data analytics in 2011-2013 and the full deployment is expected in over ten years [1].
Gartner also predicts that the aggregated value and economic benefit of the IoT will exceed $1.9
trillion in the year 2020 alone [3]. On the other side, McKinsey Global Institute, another business and
economics research firm, predicts that IoT global economic benefits to be up to $11 trillion in 2025
and business-to-business applications will account for 70% of the revenue [4].
Growth trend and forecast for connected devices is shown in figure 5, where Internet of things
occupies a big chunk of the market [8].
Fig. 5. Growth and forecast for connected devices
Increasing interest for IoT development increases the demand for lots of IoT developers as shown in
figure 6 [1]. Developers are the main force behind IoT and it is believed that 17% of the world
developers are already working on IoT applications, in particular those in Asia-Pac [3].
156
Fig. 6. Number of IoT developers 2014-2020 (Source: VisionMobile estimate 2014)
Whether these developers will be able to keep up with the demand is a major challenge. Systems need
to be designed to accommodate and provide interoperability between existing machine-to-machine
systems, different device manufacturers, and pre-existing service providers.
5. IoT Challenges Faced By The Industry
Despite the huge interest and demand for this technology and the efforts the industry is making, there
are still lots of challenges and obstacles ahead that developers face towards the full IoT deployment.
According to Harbor Research, an international research, technology and business development
consulting firm, the biggest challenge IoT developers face is to enable a seamless interoperability
between each IoT connection [4].
Another major challenge is to understand how the IoT possibilities fit into a working business model.
Challenges are also present in the development phase of the applications themselves and vendors
seem to have different views on the level of difficulties when it comes to development of these apps
and according to them location-based and monitoring apps were listed as among the most difficult
ones. In terms of best operating systems to work with IoT apps, vendors still seem to be divided and
not fully synchronized. In this regard, Android is ahead with 29% of the respondents followed by
Windows with 24% and as a platform, Java is the absolute winner with 55% [4].
Other challenges are mainly to do with the legal and logistic difficulties to connect things and devices,
provide the required intelligence at the edge and make use of the data for better decisions.
Legal and logistic difficulties can be regulatory requirements of different cities, harsh conditions, long
distance between the devices etc.
Strict requirements for reliable, manageable and secure connected things as it is the case for smart
grid is also a challenge.
6. Issues Of IoT Technology
Main concern that the industry and users have when it comes to IoT is the security issue. Security
issue has not been addressed fully yet and it is inevitable that early IoT implementations will be
insecure which will force the companies to enhance the security later on. According to HP Security
Research, about 70% percent of Internet-connected devices are vulnerable to attacks caused by
insecure Web interfaces and a lack of encryption [10]. IoT will be only as secure as the least secure
device on the network, which means that enterprises will have to ensure that every part of any IoT
they are involved with is secure and standards compliant.
Another big issue in IoT is privacy and data protection where extra efforts should be made to avoid
security breaches that would result in heavy fines and penalties. Regardless of the security in place
and given the huge data traffic travelling though IoT networks, enterprises will have specifically to
add more security to data centre infrastructure that stores and processes the data from IoT devices.
157
To address the security issue appropriately, a proper security is required at every level of the
infrastructure, i.e. security at IoT endpoints, Gateways, Networks and on the Cloud.
Fig. 7. Main elements of an IoT infrastructure where security has to apply
6.1 End Point Security:
Security on this level should ensure the following

Protect operational platform

Protect existing and new applications

Protect privacy of operators and data
6.2 Gateway Security for IoT:




Should ensure secure connectivity of devices to Cloud
Should manage security of existing devices
Should manage the data flow and access of connected devices
Should support the security of edge analytics
6.3 Network Security



Should manage security across network
Integrate security policies with Gateway and Cloud
Provide intelligent and secure traffic shaping and prioritization
6.4 Cloud and Data Centre Security



Support security and management for cloud
Protect infrastructure platforms
Secure data access between applications and operators
6.5 Other identified issues that IoT is faced with are as follows:





Lack of access to Internet everywhere, though this is improving everyday
Number of IoT sensors connected to Internet still not high enough to justify the full
implementation
Lack of standards, in particular lack of standardised communication protocols for the
sensors to avoid data silos
Lack of specified APIs to accommodate all sensor data types
Powering a huge number of lower-power gadgets without batteries
7. Standards Activities Around IoT
Connecting everything and making these ‘things’ talking to each other is very complex and difficult
task given the huge number of them and as such requires some order in this technology. Efforts to
bring some order in IoT started in late 2013 and continued in 2014 and 2015 and most likely will
continue till 2017.
One way to bring some order to this technology is through the development and adoption of open
standards which involve agreements among the key players for a common IoT architecture and
infrastructure, as shown in the figure below.
158
Fig. 8. IoT Reference Model (Source: ITU-T)
This common architectural framework will enable the industry to clearly define relationships among
the IoT’s numerous vertical markets such as transportation, healthcare, and entertainment. This will
prevent or minimize fragmentation of industry and vertical markets, will improve interoperability and
enable the industry to build IoT ecosystems to leverage the power of those connected “things” in an
efficient way.
In addition, standards enable companies to speed up their time to market and increase their market
footprint.
The complexity of another structured relationship between IoT and other technologies is shown in
figure 9 as part of a heterogeneous standards environment in IoT.
Fig. 9. Heterogeneous standards environment in IoT (Source: IoT Research EU)
Usually the pace of standards development is slower than the pace that vendors would like and
therefore this makes vendors and the industry in general form de facto standards and not wait for
formal standards bodies to complete the specifications which in the case of IoT will not finish before
159
2017. This has become very common lately for various groups to form industry alliances where
members accelerate the progress which leads to transferring to global standards. Cases like this in the
past were ZigBee, HomePlug, Wi-Fi, IEEE 2030.5 etc., that started as industry alliances and became
global standards at the end.
A number of standardization projects for IoT have been launched recently and their focus and goals
could be summarized as following:

Developing an architecture reference model to allow seamless integration of heterogeneous
IoT technologies

Develop technical specification for a common service layer to allow connectivity between
IoT devices

Develop an open universal IoT software framework

Develop intelligent industrial automation

Develop a thin interoperability layer to make data available for others to use it

To provide an open source implementation

Develop a standard interoperability architecture for information sharing

Develop an open source protocol to run over 802.15.4 to support home products

Develop a light-weight M2M protocol for remote management of M2M devices
A list of key IoT standardisation projects is given below, but there are many more that make the IoT
race more crowded and more complex.
AllSeen
Alliance
(December 2013)
This alliance came up with
an
open-source
framework,
AllJoyn,
based on a technology
from Qualcomm [5]
Qualcomm,
Cisco
Systems, Panasonic and
some other electronics
vendors
Open
Interconnect
Consortium (OIC) (July
2014)
Aims to develop standards
and certification for
devices involved in IoT.
Developed an open source
software framework,
IoTivity that enables
seamless device-to-device
connectivity to address the
needs of IoT.
Intel, Samsung, Dell, HP,
Lenovo, Broadcom etc.
Thread Group (2014)
Focuses on home security
and low-power features
that make it better for
connecting
household
devices
than
other
technologies such as Wifi,
NFC,
Bluetooth
or
ZigBee. This protocol was
initially developed by
Nest, which was bought by
Google in 2014 [5].
Google,
Samsung
Electronics,
ARM
Holdings,
Freescale
Semiconductor, Silicon
Labs, Big Ass Fans, and
Yale, a lock company
Industrial
Internet
Consortium (IIC) (March
2014)
Aims
to
accelerate
development and adoption
of intelligent industrial
automation for public use
cases
General Electric, Cisco
Systems, IBM, Intel and
AT&T,
Microsoft,
Samsung, Huawei
160
IEEE P2413 (2014)
ITU-T SG20 on IoT and
its applications including
smart
cities
and
communities
OMA LWM2M
WiFi alliance
IoT-A (2010-2013)
oneM2M (2012)
Aims to create a standard
interoperability
architecture and define
commonly
understood
data
objects,
for
information sharing across
IoT
systems;
Standardization targeted
by 2016 [6]
Aims to promote a unified
approach for development
of technical standards
including M2M (machine
to
machine)
and
ubiquitous
sensor
networks enabling the
service providers globally
to offer services supported
by this technology [6].
IEEE; collaborating with
oneM2M, ETSI and other
SDOs to evolve joint
standards
Working on a new Lightweight M2M protocol
standard,
for
remote
management of M2M
devices and related service
enablement
Aims
to
develop
certifications that align
with Internet of Things
(IoT) needs and use cases
with multiple IoT industry
input.
Developed
an
architectural
reference
model to allow seamless
integration
of
heterogeneous
IoT
technologies
into
a
coherent architecture to
realize ‘Internet of Things’
rather than ‘Intranet of
Things’
Aims to develop technical
specifications
for
a
common M2M Service
Layer
to
allow
connectivity
between
devices and various M2M
applications, to realize
horizontally
integrated
Internet-of-Things
OMA members
ITU-T members
WiFi Alliance members
ALU, Hitachi, IBM, NEC,
NXP, SAP, Siemens, and
universities – “Mission
Accomplished late 2013”
Leading ICT standards
bodies namely ETSI,
ARIB, TTC, ATIS, TIA,
CCSA and TTA
161
HomePlug Alliance (Apr
2014)
HyperCat (May 2014)
Aims
to
develop
technology specs for
powerline networking to
enable home connectivity
Aims to develop an open
specification for IoT that
will make data available in
a way that others could
make use of it, through a
thin interoperability layer.
AMD, 3Com, Cisco, Intel,
Intellon,
Texas
Instruments,
Motorola,
Panasonic at al
ARM, BT, IBM, Intel,
Living PlanIT, et al
Table 1. Main IoT standardisation projects
8. Use cases and implementation landscape
There are so many practical applications and implementations of IoT in the real world and this section
will only highlight some of them by putting them in main vertical areas of adoption [8]:

Connected Wearable Devices

Connected Cars

Connected Homes

Connected Cities, and

Industrial Internet.
Fig. 10. The IoT landscape (Source: Goldman Sachs GIR)
8.1 Connected Wearable Devices
IoT has found wide application in eHealth industry through wearable devices that help to monitor the
conditions of patients and also alert them on a potential risk.
162
Fig. 11. Connected Wearables
8.2 Connected cars
Cars are expected to be a major element of Internet of Things, where about 20 % of all vehicles, or a
quarter of a billion cars on global roads, will have some sort of wireless network connection by 2020
[15]. Use of IoT and smart devices in connected cars will increase the fuel consumption efficiency.
Fig. 12. Connected Cars
A study has shown that 13% of 2000 surveyed car owners would not consider buying a new car
without internet access, while more than 25% of them have already prioritised connectivity over
features such as engine power and fuel efficiency [15].
8.3 Connected Homes
Homes around the world are going to become smarter and more connected over the next five years,
but the demand has not reached the full potential yet.
163
Fig. 13. Connected Homes
The connected-home category will account for about 25% of shipments within the broader Internet of
Things category in 2015, but this share will increase gradually to about 27% in 2019 according to BI
Intelligence [15].
8.4 Connected Cities
Many cities in the world have started to leverage the IoT benefits to lower operating costs and
increases efficiency through automation by incorporating new technologies and capabilities through
software upgrades.
Fig. 14. Connected Cities
8.5 Industrial Internet
Industrial Internet refers to all devices, sensors, and software that enable connectivity between
machines, i.e. physical ‘thing’. Is not as broad as Internet of Things and is also known as M2M and
which describes “machines” that use network resources to communicate with remote application
infrastructure in order to monitor and control the machine itself or the surrounding environment.
164
Conclusion
This paper analysed and looked at the IoT technology as a set of enabling technologies from the
market, technology and standards perspective with the focus on drivers and hindering factors of the
full IoT deployment. It was pointed out in this paper that IoT is changing the world and this
transformation has already begun; this transformation will only continue to speed up.
From the market perspective, the key vertical markets were identified and the impact of IoT will have
in this markets was analysed.
On the technology side, different technologies were mentioned and the importance of them for IoT
was given. This paper also highlighted the importance of standardization for IoT, which provides
interoperability, compatibility, reliability, and effective operations on a global scale. The emphasise
was put on the crowd of IoT standards projects and on the list of key players in this race.
Challenges and issues were also analysed with the focus on the security of IoT and finally some
implementation cases were described.
References
1.
Asay,
M.:
Why
The
Internet
Of
Things
Is
Still
Roadblocked,
http://readwrite.com/2014/08/04/internet-of-things-obstacles-roadblocks, August 2014
2. van Rijmenam, M.: The Great Sensor-Era: Brontobytes Will Change Society,
https://datafloq.com/read/the-great-sensor-era-brontobytes-will-change-socie/211, April 2015
3. Borne, K.: 14 Benefits and Forces That Are Driving The Internet of Things,
https://www.mapr.com/blog/14-benefits-and-forces-are-driving-internet-things#.VhWyw_lVikr,
August 2014
4. Bolton, D.: Wanted: Developers To Build The Future Of The Internet Of Things,
http://arc.applause.com/2015/06/26/internet-of-things-developer-opportunities, June 2015
5. Logvinov, O.: Open Standards Will Enable the IoT’s Growth, http://electronicdesign.com/iot/openstandards-will-enable-iot-s-growth, September 2014
6. ITU-T Study Group 20: Internet of Things (IoT) and its applications including smart cities and
communities (SC&C), http://www.itu.int/en/ITU-T/about/groups/Pages/sg20.aspx, July 2015
7. Neagle, C.: A guide to the confusing Internet of Things standards world,
http://www.networkworld.com/article/2456421/internet-of-things/a-guide-to-the-confusing-internetof-things-standards-world.html, July 2014
8. Ronzio, J.: How The "Internet of Things" Is Transforming the Landscape of Corporate Events,
http://cramer.com/story/internet-of-things-transforming-the-landscape-of-events, 2015
9. Chamberlin, B.: Twenty Internet of Things trends to watch in 2015,
http://ibmcai.com/2015/01/27/twenty-internet-of-things-trends-to-watch-in-2015, January 2015
10. Jankowski, S., Covello, J., Ritchie, J., Costa, D.:Goldman Sachs InternationalThe Internet of
Things: Making sense of the next mega-trend, http://www.goldmansachs.com/ourthinking/pages/internet-of-things/iot-report.pdf, September 2014
11. Bingham, M.: IoT Perspectives, http://www.iotperspectives.com/connected-car.html, September
2014
12. Asay, M.: Maybe Asia-Pacific Developers Will Deliver The Internet Of Things,
http://readwrite.com/2014/07/24/internet-of-things-asia-pacific-developers-to-the-rescue, July 2014
13. IoT-A: Introduction to the Architectural Reference Model for the Internet of Things,
http://iotforum.org/wp-content/uploads/2014/09/120613-IoT-A-ARM-Book-Introduction-v7.pdf,
2015
14. Ramachandran B.: INTERNET OF THINGS – UNRAVELING TECHNOLOGY DEMANDS &
DEVELOPMENTS,
https://connectedtechnbiz.wordpress.com/2014/10/17/iot-unravelingtechnology-demands-developments,
15. Tata Consultancy services: Companies using IoT technologies increased revenue by 16% in 2014,
http://sites.tcs.com/internet-of-things/key-findings/companies-using-iot-programs-increasedrevenue, 2014
16. IEEE Standards Association: Internet of Things (IoT) Ecosystem Study,
https://standards.ieee.org/innovate/iot/iot_ecosystem_exec_summary.pdf, January 2015
165
Case Studies for a Markov Chain Approach to Analyze
Agent-Based Models
Florian Kitzler1, Martin Bicher2
1, 2
Vienna University of Technology, Institute for Analysis and Scientific Computing,
Karlsplatz 13, 1040 Vienna, Austria
{florian.kitzler1, martin.bicher2}@tuwien.ac.at
Abstract. Agent-Based Models have become a widely used tool in social sciences, health care
management and other disciplines to describe complex systems from a bottom-up perspective. Some
reasons for that are the easy understanding of Agent-Based Models, the high flexibility and the
possibility to describe heterogeneous structures. Nevertheless problems occur when it comes to
analyzing Agent-Based Models. This paper shows how to describe Agent-Based Models in a
macroscopic way as Markov Chains, using the random map representation. The focus is on the
implementation of this method for chosen examples of a Random Walk and Opinion Dynamic
Models. It is also shown how to use Markov Chain tools to analyze these models. Our case studies
imply that this method can be a powerful tool when it comes to analyzing Agent-Based Models
although some further research in practice is still necessary.
Keywords: Agent-Based Model, Markov Chain, Random Map Representation
1. Introduction
Agent-based modeling has become a widely used modeling technique and is nowadays used in
many fields such as social sciences, computer science, healthcare management and economics [1].
One of the many advantages over other modeling techniques is the high flexibility of agent-based
models. Another big merit is the possibility to describe heterogeneous structures. Especially these two
features make it possible for agent-based models to deal with huge data sources and model very
complex systems. The availability of more powerful computational tools helps to simulate such
complex models but there are still limitations when it comes to analyzing agent-based models. Very
complex agent-based models have a high number of parameters and usually a lot of them are not
known exactly. To parameterize or calibrate the model, a lot of simulation runs are necessary. This
leads to a high amount of computational resources and time. Another big issue is the validation of the
model. That means to find out if the right model was used to satisfy your needs. Hereby a problem
arises as appropriate methods hardly exist for agent-based models. The aforementioned problems
underline the need for analysis-methods for agent- based models. Before we briefly explain our
approach it is necessary to give short introductions to agent-based models and their connection to
Markov chains. The latter poses the key-tool for our analysis-method.
2. Introduction to Agent-Based Modeling and Markov Chains
In this section we will give a short introduction to agent-based models (short ABMs)
and Markov chains (short MCs). ABMs are models of systems composed of autonomous, interacting
entities called agents. These agents live together in the same environment and can change their
attributes after interacting with other agents or the environment. The collection of all attributes that
describe the agent is called individual state of the agent. In this work we are just regarding stochastic
ABMs. This means that random effects influence the simulation results. Stochastic models that are
easier to analyze are MCs. A MC is a stochastic process that satisfies the Markov property that is
also called “memorylessness” [2]. This means that the process “forgets” all about the past and the
166
further evolution only depends on the current state. We first define the state space of our stochastic
process. That is a finite set of states our process can be in. A time-discrete stochastic process is defined
as a sequence of random variables with state space. The variable always stands for the time at which
we observe our process. We define as the one-step transition probability from state at time step to
state in the next time step. The transition probabilities can be collected in a transition matrix, where
the entry in row number i and column number j correspond to. The transition matrix contains all
information about the MC and is important to analyze its transient and asymptotic behavior. After
starting in a certain state the MC evolves following the transition probabilities. We now want
to know in which state the process is when we observe it at a certain time. Given an initial distribution
vector we can calculate the distribution at time with the formula The initial distribution
is given as a row vector with being the probability that the MC starts in state at time. So the calculation
of the state distribution at time is just n times multiplying with the transition matrix. Under certain
conditions on the transition matrix a unique limit distribution exists with being the asymptotic
probability for the process to be in state. Under these conditions the limit distribution can be
found solving a system of linear equations, see [2]. As intended by the mentioned formulas MCs are
quite simply to analyze in the contrary to ABMs. Hence a direct comparison between these techniques
would improve the analysability of the ABMs.
3. Agent-Based Models as Markov Chains
This section will show that this comparison is possible as we give a step by step instruction on how
to create a MC from a given ABM [3], [4].
3.1 Step1: Identifying the State Space
The first step is to find all possible states of the MC. Therefore we start with a configuration of the
ABM. That is a possible combination of individual state values for each agent.
Fig. 1. Two configurations that map on the same Markov state. The first number represents the
number of black agents, the second number represents the number of white agents. The state
space of the MC is given as with number of black agents number of all white agents.
Let be the individual states space of an agent and the total number of agents. We define the
configuration space as the set of all configurations of the ABM. Our aim is the macroscopic
analysis of the model. Hence we need to find an aggregation mapping from our configuration space
to a space. Usually this is done defining that the i-th row of counts all agents with state. The image
will furthermore pose for the state space of our MC. This idea is illustrated in Fig. 1 for an ABM
with only two individual states represented by the colours white and black.
3.2 Step 2: Calculate the Transition Matrix
The second step is to find all possible transitions between our Markov states and calculate the
transition probabilities. Fig. 2 shows the practice of how to find possible transitions.
167
Fig. 2. Schematic description of the approach of calculating the transition matrix.
We first start with a chosen Markov state and we want to know which state can be reached in the
next time step. Then we select a configuration of the ABM that maps on our chosen Markov state. It
has to be mentioned that it is generally not trivial what configuration has to be chosen if more than
one of them maps on the same Markov state but in the investigated examples this is irrelevant.
With the regarded configuration we simulate one time step of the ABM using the updating rule. The
updating rule contains all information about the ABM for example the movement of the agents or
interacting rules. All agents update their attributes according to that rule and we observe a new
configuration after the time step.
The last step is to map our new configuration again onto the state space with and we have found
one possible state transition. If we repeat this several times for each Markov state we can
approximately calculate the transition matrix for our MC.
Following these two steps a MC can be developed matching the macroscopic dynamics of the ABM.
Hence the ABM can be investigated analysing the MC.
4. Results
We finally want to compare some results of the ABM and the corresponding MC for some test
cases. Therefore we look at the transient behavior of the model at a specific time. We fix an initial
distribution and calculate the transient distribution for the MC. For the ABM we realize a MonteCarlo-Simulation where the start values for the agents follow the same initial distribution.
4.1 1D Random Walk
We will take a look at the results of a 1D random walk model. In this case we have a 1-dimensional
lattice with a finite number of sites. Furthermore a number of agents are distributed on them. Hence
the only attribute of an agent is its site on the lattice. The agent can move left and right following
some rules. If the agent is alone on a site he has to move on to one of the neighboring sites with the
same probability. If another agent is on the same site, there is a probability of 0.6 to stay. First we
need to define the states of the MC and how the map works. We are considering 5 sites and 2
agents. The individual state of an agent is just an integer between 1 and 5 that holds his position on
the numbered lattice. A state of the MC always contains 5 positive integers that sum up to the total
number of agents, in our case to 2. If we use the symmetry of this model we can reduce the total
number of Markov states to only 9. In Fig. 3 we can see the so-called transition-diagram of our MC.
The arrows show all possible one-step transitions with probability greater than 0 between the
Markov states represented by circles. Without knowing the exact transition probabilities we can start
analyzing our process.
168
Fig. 3. Transition-diagram of a 1D random walk with 2 agents that can move on 5 sites lattice. The
circles represent the Markov states and the arrows represent all possible one-step transitions. On
the right hand side we see the corresponding agent configurations.
We call a closed communicating class. Once the process enters one of those states, it stays in the class
forever. As is reachable from every other state in a finite number of time steps we neglect all other
states for the asymptotic behavior.
Fig. 4. Comparison of the ABM (bright) with the time-discrete MC (dark). The process always
starts in state 5. On the left we see the transient distribution after 10 time steps, on the right
after 20 time steps. For the Monte-Carlo-Simulation of the ABM we used 100.000 iterations.
The results in Fig. 4 show that the limit distribution is nearly reached after just 20 time steps. In this
case we could calculate the transition probabilities by hand using the movement rules of the random
walk. In our simulation the stochastic process always started in state 5 corresponding to the
configuration. The exact limit distribution can be calculated solving a system of linear equations. It
is independent of the initial distribution with, and. We can also calculate the expected value of the
time we first enter the closed communicating class. When we start in state 5 this absorbing time is
reached after 7.08 time steps.
4.2 Opinion Dynamics Model
Another type of model we were investigating is a two-state opinion dynamics models in which the
agents are able to change their states after interacting with other agents. We were looking at
different interaction network types and compared the results with regard to the Euclidean norm of
the error vector. Analyzing a simulation run of the ABM with 20 agents we received a total
difference to the Markov chain of.
Conclusion and Outlook
Our case studies show that this approach works well for the investigated examples. Using statistic
tests we can show that the results of the ABM and the MC follow the same distribution. The
movement and interacting models we considered can be used in more applied models as sub models.
Some further research on bigger and more applied models is still needed. Acknowledgments. KProjekt DEXHELPP is supported by BMVIT, BMWFW and the state of Vienna via COMET –
Competence Centers for Excellent Technologies. Programme COMET is processed by FFG.
169
References
1. Axelrod, R.: Agent-Based Modeling as a Bridge between Disciplines. In: Tesfatsion, L., Judd,
K.L. (eds.): Handbook of Computational Economics, Agent-Based Computational Economics,
Volume 2, 1565-1584
2. Serfozo, R.: Probability and its Applications: Basics of Applied Stochastic Processes.
3. Banish, S., Lima, R., Araújo, T.: Agent Based Models and Opinion Dynamics as Markov
Chains
4. Izquierdo, L. R., Izquierdo, S. S., Galán, J. M., Santos, J. I.: Techniques to Understand Computer
Simulations: Markov Chain Analysis. In: Journal of Artificial Societies and Social Simulation
12(1)6 http://jasss.soc.surrey.ac.uk/12/1/6.html, (2009)
170
Hybrid Modelling in System Simulation
Andreas Körner
Vienna University of Technology,
Institute for Analysis and Scientific Computing,
Wiedner Hauptstraße 8-10, 1040 Vienna, Austria
andreas.koerner@tuwien.ac.at
Abstract. In times of increasing power and capacity in computer simulation mathematical models are
getting more and more important. For different technical applications and in natural science but also
in economic systems and management processes appropriate mathematical model descriptions are
necessary. Hybrid modelling is a special technique for more complex model descriptions in order to
reduce the degree of complexity. In different fields of interest the behavior of a model is dependent
on the active state. When the model description is changing from one state to another a so-called state
event takes place. State event modelling is the overall term to describe this modelling approach. One
state is defined by one dynamic system description and another state is described by the next
description. The mathematical environment of the model allows finding the description which is the
best matching one in each state. In this sense it is possible to find the most efficient model description
for each state and it is not necessary to build up a complicated model structure to cover all cases in
one model. Beside the principle effect of the basic structure it is also possible to combine different
mathematical modelling techniques for realizing a hybrid model for a certain complex system. In each
case a certain mathematical model of the mathematical method can be provided. This is the formal
mathematical definition of a multi method approach. In different states different models are simulated
and a certain master algorithm is managing the overall administration of the discrimination. The
mentioned issues are covered under the overall term hybrid modelling and will be introduced in the
corresponding paper.
Keywords: Mathematical Modelling, Hybrid Modelling, System Simulation
1. Introduction
System simulation is an often used term in the current scientific fields and is as well used for business
solutions. Fraunhofer ITWM refers system simulation in the context of simulation models in the
automobile sector, see [1]. In this case the term explains the integration of physical attributes in
different stages of the development process of a car. In mechatronics the term system simulation
covers the field of a 1-D simulation of multidisciplinary systems, as addressed in [2].
In this publication, system simulation is understand as a simulation approach which focuses on the
description and simulation of a whole system, e.g. a physical, electrical, mechanical,
mechatronical system or process surveyed over time. In this class of simulation approach spatial
distributed parameter simulation is not covered, neither diffusion nor relaxation processes nor similar
processes simulated by FEM methods or similar.
2. Modelling Structure and Placement in System Simulation
2.1 Hybrid System Modelling
The term hybrid modelling is within the last 20 years one of the most fashionable terms in the field
of modelling and simulation. The meaning of the term is as multifunctional as the fields of science
which uses the term in the discipline. According to title of this paper the term hybrid modelling is
171
oriented to system simulation and due to this based on dynamic systems. One basic work on dynamic
systems and the correlation to hybrid dynamic systems, their modeling and analysis of model is [3].
The basic idea of hybrid modelling is a partition of the whole model in sub-model descriptions which
are either partitioned regarding the modelling approach or the modelling technique. With respect to
the implantation in a certain simulation environment the model description can be separated for
different special cases, where the sum of the special cases are representing the model of the whole
system. Benefit of this modeling structure is the cost in the simulation environment.
2.2 Hybrid Modelling in System Simulation
System simulation is based on the description of dynamic systems. The solution of this systems are
represented by a processing over time. Hybrid dynamic systems are represented by partitioned in
different sections where the solution is computed from a sub-model description. This approach is
illustrated in Figure 1.
Fig. 1. Illustration of the sequential processing of different sub-model descriptions which performs
in total the model processing of a certain system.
One important aspect is the appearance of certain sub-models. At first consideration it
can be expected, that a certain number of models are consecutively processed up to the final model of
a simulation run, but this is wrong. The model description is partitioned in different sub-models, but
the processing over time can allocate one model more often, if the condition for transition is permitting
a certain number of requests. Figure 2 is illustrating the relation between the different sub- model
descriptions and the transition in between of those.
Fig. 2. Illustration of an exemplary Interaction of a Hybrid Model with three different
Sub-model Descriptions.
172
Figure 1 is representing a time oriented processing domain, whereas Figure 2 focus on the causal
relation between the different sub-model descriptions of a hybrid model.
The splitting of the model description in sub-model components is for simulation aspects very
important. This offers consideration in distributing different sub-models in different simulation
environments which opens the research field of co-simulation and related topics.
3. Mathematical Aspects and Formalization
The given structural description in the section before requires a detailed definition of the mathematical
model description. Mathematical aspects of the model description are the following:
• Combination of discrete states and continuous dynamic system description
• Discrete Formalization between sub-models: Finite Automaton
•
Continuous Formalization of a certain sub-model: Dynamic System - Differential (Algebraic)
Equation
•
Transition between sub-models: Jumps and Guard region combine the discrete and continuous
formalization and link the sub-model to a hybrid model
A mathematical characterization is important to have two interests: To be able to relate the separation
of the sub-models w.r.t. the overall model on the one side and the possibility to establish a layer where
the comparison of different hybrid model descriptions is possible.
The mathematical formalization of this framework in detail is listed in [4]. This current publication
addresses the purpose of hybrid models in system simulation to be able to distinguish between a
mathematical model and the simulation model in a specific simulation environment. The model per
se covers the abstract description without some considerations of the simulation environment behind.
The simulation model consider restriction of the environment and include the first attributes of the
numerical simulation.
Summary, Conclusion and Outlook
The hybrid modelling approach is offering a more efficient way to develop models for complex
systems. Of course an overall model description of the whole complex model is in most of the cases
available, but for a more efficient model description w.r.t. to simulation aspects the hybrid approach
is more powerful.
Simulation environments often restrict the possibilities of modelling and include the considerations
regarding numerical mathematics which is interesting for the simulation run of the model but not for
the model description. The presented mathematical framework offers a wide range of continuing
considerations. On the one side several considerations in the field of the simulation model can be
addresses, e.g. co-simulation and multi-method simulation methods, and plural approaches regarding
technique in the modelling method allows a multi- method approach in the abstract mathematical
environment.
References
1. Fraunhofer ITWM, http://www.itwm.fraunhofer.de, January 2016
2. Siemens PLM, http://www.plm.automation.siemens.com, January 2016
3. Lygeros J., Tomlin C., Sastry S.: Hybrid Systems: Modeling, Analysis and Control, EECS
Instructional Support Group, University of California, Berkeley (2008)
4. Körner A., Breitenecker F.: Approaches for State Event Handling by Simulation Algorithm and
via Model Description: 22. Symposium Simulationstechnik, HTW Berlin; 03.05.09.2014; in: Proceedings ASIM 2014 22 Symposium Simulationstechnik, Wittmann J., Deatcu C.
(Eds.); ARGESIM/ASIM, 43/2/Vienna, ISBN: 978-3-901608-44-5; p. 219 – 224
(2014).
173
Definition, Validation and Comparison of Two Population
Models for Austria
Martin Bicher1,2, Barbara Glock2, Florian Miksch2, Niki Popper1,2
Günter Schneckenreither1
1
Institute for Analysis and Scientific Computing, TU Wien, Wiedner Hauptstraße 8-10,
1040 Vienna, Austria
2
dwh simulation service, dwh GmbH, Neustiftgasse 57-59, 1070 Vienna, Austria
{martin.bicher, niki.popper, guenter.schneckenreither}@tuwien,ac.at
{barbara.glock, florian.miksch}@dwh.at
Abstract. In this work we present two structurally different mathematical models for the prognostic
simulation of Austria’s population: A time-continuous, macroscopic system dynamics approach and
a time-discrete, microscopic agent-based approach. Both models were developed as case studies of a
series of population concepts in order to support models for decision-support in Austria’s health care
system. In the present work we want to focus on the definition, the parametrisation as well as
especially the validation process of both population-models. The latter was of special interest as it
included a cross-model validation with Statistics Austria’s own prognostic model SIKURS.
Keywords: population model, model comparison, validation, cross-model validation
1. Introduction
As the patient always poses the centre of interest, valid prognostic modelling for decision support in
the health care system is only possible if the underlying population is predicted validly as well.
Doubtlessly long-term epidemiological or health-technology-assessment models (for example [5, 9])
can never be valid if the underlying population growth or decay is not considered.
Therefore developing models for predictive simulation of a country’s population – so called
population models – is one of the key tasks of health-care research project DEXHELPP. In order to
create a valid founding for decision-support models for Austria’s health care system a series of
quality-assured population model concepts have been researched, developed and documented. In
order to decide which of these concepts can be applied as a solid fundament of a decision-support
model, two different (very basic) population models have been implemented and validated as well.
These two case studies, a macroscopic system-dynamics model and a microscopic agent-based model,
will be focus of this work. We will explain the two model concepts and finally focus on their
interesting validation process.
2. Modelling Approaches
In this section we want to roughly describe the two used modelling methods and give a short definition
of the developed models.
2.1 System Dynamics and the System Dynamics Model
System Dynamics. System Dynamics (short SD) is a macroscopic modelling technique developed by
J.W. Forrester during the mid- 1950s [1, 2]. Herein a dynamic system is simulated using causal
relationships between its components identified by the modeller. In case the causal relationships (also
called flows) and their quantities are determined the temporal development of the system-components
(also called stocks) are simulated by ordinary differential equations.
174
System Dynamics Model. The model consists of 190 compartments, each of them representing the
number of people of a certain age from 0 to 95+ and their sex. Direct flows between them can be
interpreted as ageing. Additional flows to sinks and from sources simulate births, deaths, and
migration. The model was furthermore implemented using the simulation environment Any Logic
[3].
2.2 Agent-Based Modelling and the Agent Based Model
Agent-Based Modelling. Compared to system dynamics, agent based modelling (short ABM) is a
very young modelling technique as it became popular among modellers during the 1990s. It models
individuals as agents with individual attributes and behavior. Classically these models are simulated
with equidistant time-steps wherein each agent is addressed once to perform some behaviourcorresponding actions. As there is no definition of the modelling approach, all scientific fields agree
with, the reader is referred to modelling guidelines presented at the 2006 Wintersimulation
Conference [4] for more detailed information. Agent-Based Model. Each model agent represents one
person in reality. Hence it is given a certain sex and age wherein the latter enhances with time.
Additionally each agent has age- and sex-dependent probabilities to die and to emigrate. In case the
agent is furthermore female, there is a probability to give birth to a child as well. Due to immigration,
a number of individuals are additionally generated. The whole system is then simulated with time
steps of arbitrary length (classically between 1 and 365 days). As Austria’s population (about 8 Mio
people) is unfortunately too big in order to simulate each individual at once, one model-agent is
defined to pose for 100 persons. The model was moreover developed using the open source
programming language Python 3 and executed with CPython.3.3.
3. Parameterization and Validation
3.1 Concepts
In order to correctly simulate the two models a specific research question is fixed: “Simulation of
Austria’s population from 01.01.2003 to 31.12.2050 with respect to age and sex”.To reach this target
both of the models first of all need to be correctly parameterised. Therefor datasets gained from
Statistics Austria, probably Austria’s biggest base for open-source socio-economic data and statistics
[8], were used. Hereby the so called STATcube [7] provides a user friendly tool to search for data
with almost arbitrary fineness and filters.
Altogether data-sets for migration, fertility and deaths for 2003-2014 were collected, each of them
broken down to yearly age-categories and sex. Furthermore Statistics Austria applies the prognostic
tool SIKURS [6] to create prognosis data up to the year 2076. Hence we additionally inquired
prognostic data-sets according to the aforementioned real-data. As fertility prognoses were only
available on the average, we developed a statistical extrapolation model to obtain this data for different
age- categories as well. This data finally posed a firm basis for a direct parametrisation of both models.
Hereby different time-scaling formulas have to be applied as the SD approach is time- continuous
using transition-rates while the ABM is time-discrete with equidistant steps using transitionprobabilities. The target that the ABM should be valid for arbitrary length of the time-steps poses an
additional challenge. As Statistics Austria additionally provides standard population data and
prognosis also a basis for the validation-process of both models is provided: First, all models’ results
can be compared with collected population data for 2003-2014. Second, all type of cross-model
validation can be performed comparing the models’ results with prognostic data gained from the
SIKURS tool via Statistics Austria. The total parametrisation and validation concept is visualized in
Fig. 2.
175
Fig. 2. Parametrisation- and validation-concept of both population models.
3.1 Validation Results
After brief verification and basic face-validation, finally, empirical validation was performed for both
models at the same time. We decided to compare Statistics Austria (prognostic) data with the
simulation results of the SD model and the ABM for two different time-step sizes (30 days and 365
days). As the ABM is a stochastic model,
50 simulation runs were executed and arithmetically averaged for the comparison.
Fig. 3 shows a direct comparison of the total population. Here we want to state a possible
interpretation of three remarkable observations.
1. First, the SD results curve perfectly fits the prognosis curve of Statistics Austria. Our interpretation
of this feature is that the prognostic SIKURS tool of Statistics Austria works quite similar to our
system-dynamics model.
2. Second, it seems that the ABM with smaller step-size (30 days) notably overshoots the prognosis
curve of Statistics Austria.
3. Third, this ABM simulation can be seen to perfectly fit the real-data between 2003 and 2015 while
the SD model and the ABM with 365 day steps fail.
Thus, in terms of validity, we are facing the problem that the model producing better results in
comparison with the real data reference, drifts off the reference for prognostic values.
Differences between different time-intervals of the ABM occur due to the parameterisation process
respectively the parameter-calculation: Due to feedback reasons it is impossible to find an analytical
formula to correctly scale the time-step- dependent transition-probabilities of the ABM.
176
Fig. 3. Comparison of the total population for both models’ results and Statistics Austria
prognosis and collected data.
In addition to the empirical comparison of the total population several other tests were applied for
both models each affirming the validity of both models. We furthermore want to lay special emphasis
on the comparison of the age-structure of the population. Hereby we compared the age-pyramid for
different points in time. The result can be seen in Fig. 4 on the example of 01.01.2015.
We see that the AB model perfectly fits the actual demographics of 2015. The SD model produces a
smooth demographic curve, which is a natural artefact of SD models and can be considered as
acceptable.
Conclusion
Both modelling approaches can be seen to be valid with respect to the data gained from Statistics
Austria. In terms of different step-sizes, the agent-based approach leads to slightly different results. It
finally depends on the availability and quality of data, which of them can be seen to be more correct.
The SD approach perfectly matches the Statistics Austria prognosis for the aggregated numbers, but
some flattening effect can be observed regarding demography. Furthermore the SD model stands out
due to its fast computation time while the ABM provides a more flexible structure for extensions like
regionality or socio-demographic factors. Summarising both models can be seen to validly simulate
Austria’s population and might be extended to simulate diseases and possible interventions.
177
Fig. 4. A comparison between both models and Statistics Austria data from 01.01.2015 for different
age-pyramids is shown.
Acknowledgments. K-Projekt DEXHELPP is supported by BMVIT, BMWFW and the state of
Vienna via COMET - Competence Centers for Excellent Technolo-gies. Programme COMET is
processed by FFG.
References
1. Forrester, J.W.: Industrial dynamics. Productivity Press, Cambridge, MA (1961).
2. Forrester, J.W.: World Dynamics. Cambridge, MA: Wright-Allen Press (1971).
3. Grigoryev, I.: AnyLogic 6 in three days: a quick course in simulation modeling. AnyLogic
North America, Hampton, NJ (2012).
4. Macal, C.M., North, and M.J.: Tutorial on Agent-Based Modeling and Simulation Part 2: How to
Model with Agents. In: Proceedings of the 2006 Winter Simulation Conference. pp. 73–
83, Monterey, California (2006).
5. Miksch, F. et al.: Modelling Spread of Pneumococcal Diseases in Austria: Long Term
Behavior and Impact of Vaccination. In: Elst, G. (ed.) ASIM-Workshop 2009 in Dresden mit
integrierter DASS’2009 - Workshop Beiträge. pp. 147–152 Fraunhofer IRB Verlag, Dresden,
Germany (2009).
6. Statistik Austria: Demographisches Jahrbuch Österreich 2012. Verlag Österreich GmbH
(2012).
7. Statistik Austria: StatCube,
http://www.statistik.at/web\_de/services/datenbank\_superstar/020627.html.
8. Statistik Austria: Statistik Austria, www.statistik.at.
9. Urach, C. et al.: Model selection process for cost effectiveness analysis of an organized
AAA screening in austria. In: ISPOR 15th Annual European congress. p. A473 (2012).
178
Microsimulation Models for Simulating Pathways of
Patients with Mental Diseases
Andreas Bauer1, Felix Breitenecker1, Christoph Urach2
1 TU Wien, Institute for Analysis and Scientific Computing, Wiedner Hauptstraße 810, Vienna, Austria.
2 dwh simulation services, Neustiftgasse 57-59, Vienna, Austria.
andreas.e101.bauer@tuwien.ac.at
Abstract. Predicting demand for care is necessary to provide sufficient capacities in hospitals and
enable better planning to cope with challenges like changes in the structure of the population.
Emphasis in this work is put on the analysis of the data and on the parametrization of a simulation
model for the paths of patients with mental diseases through the health care system. Survival analysis
and model selection methods are used for this purpose. Data on patients and their treatments is
analyzed with methods of survival analysis. Different methods for modelling the survival and hazard
function are presented and compared. Hereby, the focus is on the cox model. It is used to model the
hazard function and can be extended to analyze multiple events. With the use of model selection
methods the significant parameters are determined. Only these are included in the simulation model.
These methods shall help to raise the quality of the parametrization and therefore the whole simulation
model. In the microsimulation model, every patient has a particular set of parameters and can be in
one of several predefined, exclusive states. The events are implemented as state changes. The
probabilities for the events are calculated using the hazard functions. These are estimated with several
extensions of the cox model. The considered events in the simulation are readmissions to hospital and
contacts to ambulant psychiatrists. The simulation runs for a predefined time span and the sequence
of events of each patient is tracked. After the simulation, the individual paths of the patients as well
as aggregated quantities such as the overall numbers of certain events are analyzed. Simulations for
different populations are performed. The results for various scenarios are presented. Scenarios with
and without contacts to a psychiatrist are considered as well as different maximum numbers of
admissions. Also, the subpopulations are compared. For example, differences in the results for
diagnosis groups are encountered. These simulations shall lead to an improvement of the prediction
of the pathways of the patients and therefore help to evaluate interventions like treatment changes the
health care system or the utilization of the capacities in hospitals.
Keywords: Microsimulation, Survival analysis, Cox regression
1. Introduction
Predicting demand for care is necessary to provide sufficient capacities in hospitals and enable better
planning to cope with challenges like changes in the structure of the population. This work deals
particularly with patients with mental diseases. For the purpose of simulating the demand of these
patients, a microsimulation model is built. In order to improve the quality of these simulations the
emphasis in this work is put on the analysis of the data and on the parametrization of the simulation
model for the pathways of patients with mental diseases through the health care system. Survival
analysis methods are used for this purpose. The focus is put on the Cox model and its extensions.
2. Methods
Methods from the field of survival analysis are used to build the statistical model behind the
simulation model. Survival analysis deals with the analysis of data of the time until the occurrence of
179
a particular event. This kind of data is frequently encountered in medical research and also in other
areas of application and referred as survival data. In this work the considered event are readmission
to hospital, contact to a psychiatrist and death. The methods allow the estimation and the analysis of
the survival function and the hazard function. The survival function is defined as the probability
that an individual will survive up to time and the hazard function is defined as the instantaneous rate
of death at time . The cumulative hazard function is given by. The Nelson-Aalen estimate is an
estimate for the cumulative hazard function. Let and denote the numbers of people that experience
the event at time t respectively are at risk at time t. Let denote the event times. Then, can be
estimated by
(1)
The Cox model is a model for the hazard function [1]. It assumes that the ratio of the hazards of
different exposure groups remains constant over time. This is called the proportional hazards
assumption. The mathematical form of the model is:
(2)
The baseline hazard refers to a particular group of individuals (for example, the individuals with value
zero in all categories), is the number of covariates, is the value of the th covariate and is the
corresponding regression coefficients.
An extension of the Cox model is the stratified Cox model [2]. This model allows multiple strata. The
strata divide the subjects into disjoint groups and each subject is member of exactly one stratum. Each
of which has a distinct baseline hazard function but common values for the coefficient vector. The
overall log likelihood is the sum of the log likelihoods of each stratum. In the multi-state model every
transition between states is possible [2]. Every transition defines a stratum. So, the stratified Cox
model can be applied. The transition probabilities can be calculated from the cumulative transition
hazards [3]. Let be a time-inhomogeneous Markov process with state space. The cumulative
transition hazards, can be estimated through the multi-state extension of the Cox model and. This
leads to the approximation of the transition probabilities with a partition of the interval [s,t] and:
(3)
3. Model
The chosen model type is a microsimulation model. Iit follows the bottom-up approach and every
single individual is modelled. This approach is chosen because not only the cross-sectional analysis
is important but also the longitudinal pathways of individuals. This approach is suitable for analysis
of different policies and scenarios. The characteristics of the individuals are manageable with a
bottom-up approach.
In the microsimulation model, every patient has a particular set of parameters and can be in one of
several predefined, exclusive states. Patient parameters are sex, age, length of stay in the psychiatric
department of the hospital and diagnosis. The events are implemented as state changes. The possible
ways through the states are described by a transition matrix which can be interpreted as a directed
acyclic graph. The transitions are evaluated using first-order Monte Carlo simulation by comparing a
random number with the given transition probability. The probabilities for the events are calculated
using the hazard functions for the transitions. These are estimated with the multi-state model. The
considered events in the simulation are readmissions to hospital, contacts to ambulant psychiatrists
and death. The simulation runs for a predefined time span and the sequence of events of each patient
is tracked. Every individual starts in state R (released after the first admission to hospital). If the most
recent event of the patient was the i-th readmission, it is in state, if the most recent event was the i-th
ambulant psychiatrist visit, the patient is in state and if the patient died, it is in state. From initial state
transitions are possible to the states, and. From the patient can go to and from to, and to. From death
no transitions are possible since it is an absorbing state.
4. Simulations
180
Various simulations of the microsimulation model are carried out for a simulation time of two
years. Stacked area plots of the evolutions of the patient’s distribution over the states comparing
psychotic and nonpsychotic patients are presented in Figure 1.
Fig. 1. Evolution of the patients distribution over the states for psychotic and nonpsychotic patients
The percentage of psychotic patients still remaining in state is decreasing much faster than for nonpsychotic. At the end of the simulation 41% of the psychotic patients are still in state while about 55%
of the non-psychotic patients are in state. The percentages of patients in the psychiatrist states run
very similarly in both subpopulations.
Summary and Outlook
A microsimulation model for the simulation of the pathways of patients with mental diseases was
successfully implemented. The Cox model proved to be a helpful tool for the statistical model. The
main benefits of the Cox model are the inclusion of patient parameters and the flexibility for
extensions for multiple events. The first simulation results already outline a difference between
psychotic and nonpsychotic patients in terms of the number of events. Psychotic patients have
significantly more events during the simulation than nonpsychotic ones.
Work that is still open is a deeper analysis of the simulation results. Also, the definition and
comparison of different scenario will be executed and the usefulness of an intervention strategy will
be tested.
References
1. Cox, D. R.: Regression models and life-tables. Journal of the Royal Statistical Society.
Series B (Methodological) (1972) 187-220
2. Therneau, T. M., Grambsch, P. M.: Modeling survival data: extending the Cox model. Springer
Science & Business Media, 2000.
3. J. Beyersmann, J., Putter, H.: A note on computing average state occupation times.
Demographic Research (2014), vol. 30, no. 62, 1681-1696
181
International Conference on
Information Systems and
Security
183
IC-ISS
International Program Committee:


















E.Hajrizi (RKS), Chair
Breitenecker F.(AUT)
Gashi I. (UK)
Hofbauer P.(AUT)
Kopacek P.(AUT)
Mirijamdotter A. (SWE)
Schwaerzel H.(GER)
Skenderi S .(RKS)
Stapelton L.(IRL)
Yusuf S. (UK)
Baltes J.(CAN)
Qarri A. (AL)
Jesse N.(GER)
Mili F. (USA)
Wagner C. (USA)
Seiler W. (AUT )
Sev rani K. (AL)
Yayilgan S .Y. (NOR)
National Organizing Committee:








E.Hajrizi (KOS), Chair
Pireva K. (RKS)
Bazini E. (AL)
Hyso A. (AL)
Gashi B. (RKS)
Sherifi M (RKS)
Limani Y.(RKS)
Morina L. (RKS)
Editors:
Edmond Hajrizi (UBT) & Felix Breitenecker (AUT)
183
International Conference on Information Systems and Security, Nov 2015
E-learning systems in higher education institutions: An
outlook of their use in the Western Balkan Region
Blerta Abazi Chaushi1, Agron Chaushi1, Florije Ismaili2
Faculty of Business and Economics, SEE-University,
1200 Tetovo. Macedonia
2
Faculty of Contemporary Sciences, SEE-University,
1200 Tetovo, Macedonia
{b.abazi, a.caushi}@seeu.edu.mk
1
Abstract. This study provides a review of the literature on e-learning systems evolution and
environments. The argument is that e-learning systems should be embedded in the core strategy of
the institution. To support this premise, studies for e-learning are analyzed and six recommendations
are drawn for universities to follow in order to have successful e-learning environments. The main
contribution of this study, however, is the identification of the trends and statistics regarding the elearning usage in the Balkan region. These stats are identified through a survey conducted in 40
universities in 10 countries from this region. The results show that more than 70% of the universities
have adopted LMS, which does not fall short behind when compared with universities in the world.
Also, the results show that around 64% of the private universities develop LMS in-house, compared
with around 38% of the public universities, which have funding from the governments and can
purchase vendor based solutions. However, the results from the survey suggest that public universities
in these countries are more prone to open-source rather than vendor based.
Keywords: e-learning., learning management systems, higher education institutions
1. Introduction
In this study, the impact of technology in University setting is discussed, and the importance of elearning environments is explored. An analysis of the phases of evolution of e-learning and e-learning
environments is carried out. The importance of adding e-learning into the strategic plans of the
universities is presented through an examination of the literature. The result from the analysis
materializes through six recommendations that can be adopted by institutions of HE to better integrate
their e-learning systems with their strategic plans and operations. Moreover, the application of the
types of e-learning system in Balkan countries is examined. In this study 40 Universities in 10
countries (Macedonia; Albania; Bosnia and Herzegovina; Bulgaria; Croatia; Greece; Kosovo;
Montenegro; Serbia; and Slovenia) are surveyed to attain the levels of e-learning adoption and the
systems they use. This survey was conducted in year 2014. A limitation of the study is that the sample
of the universities is not large enough to generalize for the Balkan region, but since in this sample
only renowned universities from each country are surveyed, this limitation is alleviated.
2. Literature Review
The changing landscape of Higher Education Institutions (HEIs) in order to achieve thriving learning
experience and continuous improvement is being perceived through constant adjustments in the
approaches to new technology [1]. Knowledge development in the information age is a
technologically aided activity [2]. Learning Management Systems (LMS), distance education and
online learning, have become an important feature of online service delivery within the Higher
Education Information Services sector, requiring close attention to issues of functionality,
184
sustainability and usability. In today’s university landscape, a key strategic issue regarding online
learning is not whether to engage, but how [3].
2.1 Learning Management Systems evolution
Learning management systems (LMS) can be defined as: “a software application that automates the
administration, tracking, and reporting of training events” [4]. The earliest LMS can be dated back to
Sydney’s Automatic Teacher in 1924, a primitive system that automated grading of multiple choice
tests [5]. This is seen as the first endeavor to what we call today Learning Management Systems. In
1956 SAKI was invented [6], a system that automatically adjusted the difficulty of questions based
on the performance of the user. In 1969 Arpanet, the precursor of today's web was created, which will
have a huge impact in the way LMS was developed. It was until 1997 that interactive learning network
was designed. Courseinfo is amongst the first players in this sphere [7]. In year 2002 Moodle was
released, which is today one of the most used LMS in university settings in the world
(moodle.net/stats). And today, most modern LMS are hosted in the cloud, which make much easier
the process of moving to e-learning environment since there is no need to install or maintain a system
and especially no need for the burden of in-house development [8]. In the year 2005, online programs
become available at colleges and universities in different formats, as certificates, diplomas, degrees,
or post-baccalaureate programs [9]. Over the next few years, we see a slight shift from the traditional
classrooms to more network based teaching, and the beginning of the transformation of the institutions
[10]. The last decade is the time of the rapid development of technology-based learning tools.
According to a study conducted by Ingerman and Yang [11]: “This rise in strategic importance is
evidence that technology has moved beyond the data center and institutional administrative systems
and is now part of daily life for faculty and students”.
2.2
E-Learning and its integration in the strategy of the university
E-learning must be seen as a strategic initiative and operation of the institution that takes this
endeavor. Successful e-learning implementation depends on building a strategy that meets the needs
of the learners and the business goals of the institution [12]. Several advices for universities that have
adopted or plan to adopt e-learning, and especially for the institutions that want to achieve operational
effectiveness as well as increase of quality in terms of e-learning usage can be drawn. Consistent with
a research conducted by Elaine Allen and Jeff Seaman based on responses from over 2,800 Chief
Academic Officers (CAOs) and academic leaders, 69.1% of the respondents agreed that online
learning is of strategic importance for the university [13]. The reason for its strategic importance is
that the number of the students that take online courses has increased a lot in this 10 year period. It is
a fact that around 72% of universities had online offerings even ten years ago. A major change that is
worth mentioning is that a far larger proportion of higher education institutions have moved from
offering only online courses to providing complete online programs (62.4% in 2012 as compared to
34.5% in 2002) [13]. An analysis of the studies regarding the importance of e-learning in university
setting [14]–[18], as well as the studies regarding the e-learning frameworks, especially one of the
most famous, the TPACK framework [19]–[25], with a focus on the strategic role of e-learning and
importance in active learning, better teaching and creation of engaging learning environments, led to
the proposal of the following six suggestions for improvement:
1. Institutions of Higher Education should make e-learning initiatives part of the institution's
strategic plan and budget, and set specific goals for e-learning initiatives.
2. HEIs should try to centralize essential e-learning technology services as much as possible
because of the greater efficiency and seamless integration of e-learning services.
3. E-learning should be viewed as critical to the mission of the university and the provision of
e-learning services should have high priority within IT. The reliability of the technology
used in e-learning should be seamless.
4. The university should create a clear path that will demonstrate the merits of e-learning for
both traditional face-to-face and online classrooms. It is important to keep the faculty and
students interested in this new learning environment, and it is a good add-on to
implementing a faculty e-learning mentoring program using faculty who have already
taught e-learning courses.
185
5.
6.
Universities should ensure that the chosen technologies for e-learning are scalable, by
creating a plan for the number of courses/programs that will roll out e-learning initiatives
in the coming years
Universities should make sure that the chosen technologies adaptable – not all the courses
and programs have a need for the same technology.
3. Methodology
For the purpose of this study 40 universities from 10 countries in Balkan region including Macedonia,
Albania, Greece, Bulgaria, Serbia, Kosovo, Montenegro, Bosnia and Herzegovina, Croatia and
Slovenia were surveyed. The following questions: Does the university use any learning management
systems; what type of system do they use, and what specific product they use – were asked and answers
were obtained. The data was acquired by conducting a short survey that covered only the
abovementioned issues. The websites of these universities were also visited, and in cases there was
not enough information, representatives of the university were contacted by asking questions that
were needed for this study. As mentioned in the introduction of this study, the data might not be 100%
accurate since not all universities in this region were included, and this poses a limitation. However,
since this sample is consisted of renowned universities from each country, this limitation is reduced,
and the results of this study are important and reflect the reality of current state of e-learning in the
region. These figures are just a sample, and data might vary if all the universities were surveyed.
4. Results of the Survey within Universities in the Balkan Region
According to our study, about 72% of the universities in the Balkan region have some system in place
for uploading course content as lecture notes, assignments, etc. This number, in contrast with the
Universities in the world, does not stand behind very much. According to latest CDS survey, which
surveys more than 2000 universities around the world, 65% have online learning platform in place,
17% use the cloud services, 9% have some sort of a system, and only 9% of institutions do not have
discussion to date about online learning [17].
Table 4: LMS adoption level in the Western Balkan Countries
Heading level
Public
Private
Overall
Do not have an LMS
30.43%
23.53%
27.50%
Have implemented an LMS
69.57%
76.47%
72.50%
During this study, an analysis was conducted to see whether there is a difference in implementation
of LMS depending on the type (public, private) of the university. In this study 17 universities were
private and 23 universities were public. Based on the data that were collected, there is no significant
difference between and private universities on implementation of LMS. As can be seen from Table 4,
around 77% of private universities and 70% of the public universities have implemented a LMS.
Also, an interesting conclusion from this survey, as can be seen from the Figure 5: LMS Adoption
Level by Country
, is that only Greece, Bulgaria, Croatia and Montenegro have a LMS in place in all of the surveyed
universities, whereas in the rest of the countries there are still universities that are not using any LMS
as a part of regular or distance learning studies.
186
Figure 5: LMS Adoption Level by Country
Another question that this study was trying to find an answer to was to find out what type of LMS do
universities use (see Figure 6-left). The results show that, of the universities that have a Learning
Management System in place, around 43% have created their LMS in-house whereas the others have
implemented a readily available package. From the LMS packages, Moodle is definitely the most
popular one. We assume that the main determinants for choosing Moodle are that it is free, it is opensource and it has a big community. Interestingly, there can be seen a similarity in the preference of a
LMS package, with some minor differences. The major difference can be seen in the preference of
Figure 6: Types of LMS adopted Overall (left) and in Public and Private
Universities(right)
choosing a readily available solution or building the LMS in-house. 64% of the private universities
have chosen to build their own LMS versus 38% of the public universities with the same choice (see
Figure 6-right). It is worth to mention that all the analyzed public universities in Croatia use and LMS
called Studomat.
5. Findings and Conclusion
In general, the results of this study show that, Balkan region is not lacking in implementing LMS as
a technology or platform for e-learning. The tendency is to develop a LMS in-house or adopt an open
source or free LMS which is understandable due to lack of funding for IT. However, this study is
falling short in answering to the question how much are these LMSs used in reality and to what extent
do they fulfill their mission of transferring knowledge through online media. These questions need a
broader research where many universities will be involved and a better qualitative and quantitative
data will be collected to answer the questions that are addressed but not answered as a part of this
research. In summary, there is a growing recognition of the need for e-learning solutions in the Balkan
region. Institutions are becoming aware of the trends in e-learning. As it was showed in the above
section, most of the universities are making their steps in adopting LMS and some of the universities
have started with offering some distance learning and online programs. From the analysis, it was
noticed that all these universities that were part of the survey, either had open source, either build inhouse LMS. None of these universities has gone for a vendor based solution because they are in the
187
beginning stages of e-learning and they do not want to spend funds for a vendor based solutions.
However, it is encouraging to see that both, public and private universities embrace the trends in
higher education. In spite of this, we should highlight that e-learning is not only a technological
matter. Rather, it is a process involving academic staff, students, and pedagogical content. For this
reason, it is important for universities to have a strategic approach to e-learning. It takes a lot more
than providing a technological platform for e-learning in order to be successful in transferring
knowledge to students using electronic media. If higher education institutions view e-learning purely
as a technology, they will be doomed to fail [26].
References
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
L. Abazi-Bexheti, “Development of a learning content management systems,” J. Online Learn.
Teach., vol. 2, pp. 1034–1048, 2008.
D. R. Garrison, T. Anderson, and W. Archer, “A theory of critical inquiry in online distance
education,” Handb. Distance Educ., vol. 1, pp. 113–127, 2003.
S. Grajek, “Top-Ten IT Issues, 2014,” EduCause, 2014.
R. Ellis, “Field Guide to Learning Management Systems,” ASTD, 2009.
L. T. Benjamin, “A history of teaching machines.,” Am. Psychol., vol. 43, no. 9, pp. 703–712,
1988.
G. Pask, “SAKI: Twenty-five years of adaptive training into the microprocessor era,” Int. J.
Man-Mach. Stud., vol. 17, no. 1, pp. 69–74, Jul. 1982.
D. M. Fahey, “Blackboard COURSEINFO: Supplementing In-Class Teaching with the
Internet,” Hist. Comput. Rev., vol. 16, no. 1, pp. 29–37, Jan. 2000.
F. M. M. Neto and F. V. Brasileiro, Eds., Advances in Computer-Supported Learning:. IGI
Global, 2006.
B. I. Dewey and P. B. DeBlois, “Top-ten IT issues, 2006,” Educ. Rev., vol. 41, no. 3, p. 58,
2006.
B. Abazi Caushi, A. Caushi, and Z. Dika, “A Comprehensive Aproach to Technology Issues
and Challenges for Higher Education Institutions,” in 12th International conference e-Society,
Madrid, Spain, 2013, pp. 177–184.
B. L. Ingerman and C. Yang, “Top-Ten Issues, 2011.”
E. Engelbrecht, “A look at e-learning models: investigating their value for developing an elearning strategy,” Progressio, vol. 25, no. 2, p. p–38, 2003.
E. Allen and J. Seaman, Changing Course: Ten Years of Tracking Online Education in the
United States. ERIC, 2013.
M. J. Rosenberg, E-learning: Strategies for delivering knowledge in the digital age, vol. 3.
McGraw-Hill New York, 2001.
S. Alexander, “E-learning developments and experiences,” Educ. Train., vol. 43, no. 4/5, pp.
240–248, 2001.
R. H. Wild, K. A. Griggs, and T. Downing, “A framework for e-learning as a tool for
knowledge management,” Ind. Manag. Data Syst., vol. 102, no. 7, pp. 371–380, 2002.
Paramythis and S. Loidl-Reisinger, “Adaptive learning environments and e-learning
standards,” in Second European Conference on e-Learning, 2003, vol. 1, pp. 369–379.
M. Ebner, “E-Learning 2.0= e-Learning 1.0+ Web 2.0?,” in Availability, Reliability and
Security, 2007. ARES 2007. The Second International Conference on, 2007, pp. 1235–1239.
M. J. Koehler, T. S. Shin, and P. Mishra, “How do we measure TPACK? Let me count the
ways,” Educ. Technol. Teach. Knowl. Classr. Impact Res. Handb. Framew. Approaches, pp.
16–31, 2011.
Hu and V. Fyfe, “Impact of a new curriculum on pre-service teachers’ Technical, Pedagogical
and Content Knowledge (TPACK),” Curric. Technol. Transform. Unkn. Future Proc. Ascilite
Syd. 2010, 2010.
L. Archambault and K. Crippen, “Examining TPACK among K-12 online distance educators in
the United States,” Contemp. Issues Technol. Teach. Educ., vol. 9, no. 1, pp. 71–88, 2009.
Sahin, “Development of survey of technological pedagogical and content knowledge
(TPACK).,” Turk. Online J. Educ. Technol.-TOJET, vol. 10, no. 1, pp. 97–105, 2011.
M. Bower, J. G. Hedberg, and A. Kuswara, “A framework for Web 2.0 learning design,” Educ.
Media Int., vol. 47, no. 3, pp. 177–198, 2010.
188
24 E. Rahimi, J. Van den Berg, and W. Veen, “A framework for designing enhanced learning
activities in web2. 0-based Personal Learning Environments,” in World Conference on
Educational Multimedia, Hypermedia and Telecommunications, 2013, vol. 2013, pp. 2222–
2231.
25 Maor and P. Roberts, “Does the TPACK framework help to design a more engaging learning
environment?,” in World Conference on Educational Multimedia, Hypermedia and
Telecommunications, 2011, vol. 2011, pp. 3498–3504.
26 G. Ssekakubo, H. Suleman, and G. Marsden, “Issues of Adoption: Have e-Learning
Management Systems Fulfilled Their Potential in Developing Countries?,” in Proceedings of
the South African Institute of Computer Scientists and Information Technologists Conference
on Knowledge, Innovation and Leadership in a Diverse, Multidisciplinary Environment, New
York, NY, USA, 2011, pp. 231–238.
189
Credit Information System in Albania
Valbona Çinaj1, Bashkim Ruseti2
1,2Faculty
of Economic, University of Tirana
{bonaribaj1, bruseti2}@gmail.com
Abstract. The booming lending period and many lenders (16 banks and 21 non-bank financial
Institutions in Albania) brought about unprecedented competition in credit markets within Albania.
Economists usually view lending and competition favorably, but in Albania resulted in a number of
unforeseen non-performing loans. Findings report increased problems of borrower over-indebtedness,
reduced loan repayment incentives, and growing debts for lenders (Campion 2001; McIntosh and
Wydick 2005). The weakening performance of lenders is due in part to the absence of information
sharing in these markets. Because growing numbers of lenders (banks and non-bank financial
Institutions in Albania) increase the level of asymmetric information between lenders, credit
information systems (often called credit reporting bureaus or credit bureaus) can play a crucial role
towards improving credit market performance and, in turn, credit access for the poor. Increases in
formal sector lending among the poor have created a need for credit information systems that provide
potential lenders with borrower information. Improved screening affects from the system causes the
level of non-performing loans to decline.
In this paper we will present effects of a credit information system to be implemented in Albania.
Keywords: Credit information system, credit score, credit history, non –performing loans.
1. Introduction
Most of the assets of the commercial banks include different kinds of deposit which are in fact the
debts of the banks during banking and economical activities which are exposed to different forms of
risk; the most important kind is Credit Risk (CR) Financial institutions (FIs) are very important in any
economy. Their role is similar to that of blood arteries in the human body, because FIs pump financial
resources for economic growth from the depositories to where they are required Commercial banks
(CBs) are FIs and are key providers of financial information to the economy. They play even a most
critical role to emergent economies where borrowers have no access to capital markets There is
evidence that well-functioning CBs accelerate economic growth, while poorly functioning CBs
impede economic progress and exacerbate poverty The magnitude and the level of loss caused by
credit risk compared to others is severe to cause bank failures .Lending has been, and still is, the
mainstay of banking business, and this is more true to emerging economies where capital markets are
not yet well developed. To most of the transition economies, however, particular, lending activities
have been controversial and a credit risk matter. This is because business firms on one hand are
complaining about lack of credits and the excessively high standards set by banks, while CBs on the
other hand have suffered large losses on bad loans. It has been found out that in order to minimize
loan losses and so as the CR, it is essential for CBs to have an effective CRM system. Lenders
experience positive net revenue impacts from lending if they increase the classification power of their
credit scoring systems. If loan officers’ subjective assessments of otherwise intangible borrower
characteristics contain additional information about a borrower, a lender may improve the default
forecast quality of his internal credit scoring systems by utilizing this subjective information. The
Basel II regulatory framework requires lenders to use all available information about a borrower, both
subjective and non subjective, but at the same time produce consistent and objectified borrower
ratings. However, soft information is often laden with inconsistencies due to the lack of comparability
of different raters’ assessments and the existence of incentives to manipulate the soft rating. The
industry trends continued with companies that developing modeling in various applications. Why
Credit? The reason is that credit score is so important today in credit process. Facts that after a 20-
190
years history companies are going beyond credit, to build scoring models after this all experience in
Developing of Nationalization of banking and credit industries. How do companies try to stay
competitive regarding the use of credit? How do companies prepare for increasing regulatory
constraints? More diversified use of credit cards, lowering of down payments, increased risk
tolerance, Automated Underwriting Systems (AUS).The few opportunities to build credit for Families
with low or poor traditional credit. Niche market high potential returns for risky lending. Credit
scoring first emerged in the late 1950s to support lending decisions by the credit departments of large
retail stores and finance companies. By the end of the 1970s, most of the nation's largest commercial
banks, finance companies, and credit card issuers used credit-scoring systems. Over these two
decades, the primary use of credit scoring was in evaluating new applications for credit, and creditors
used their own experience and data, sometimes with the aid of consultants, to develop the models.
Although often available at the time from local credit bureaus (today more commonly referred to as
credit-reporting agencies), credit history records were limited in scope and relatively expensive to
access. Thus, lenders essentially had no practical way of obtaining the complete credit histories of
noncustomers and so could not effectively target them for solicitations on the basis of credit history.
By the late 1980s much had changed. Creditors were no longer restricted to the credit histories of their
own customers and credit applicants. Rather, lenders could purchase the generic credit history scores
of individuals who were not their account holders and, with that data, market consumer credit products
tailored to various credit scores to appropriate potential borrowers. The use of credit scoring then
spread to additional loan products including home mortgage and small-business lending. Scoring
technologies also were applied in new ways, such as in assessments by institutions of whether to
purchase individual loans or pools of loans backing securities. Finally, credit-scoring technologies
were developed to focus on outcomes beyond credit-risk assessment to include, for example, account
profitability and various aspects of account management. As the use of credit scoring was growing,
so was the demand for consumer credit and the number of credit instruments offered to finance such
activities. Since the early 1900s, merchants have been offering installment credit to allow customers
to stretch out their payments for the purchase of furniture, major appliances, and other large durable
goods. Charge cards, such as those offered by oil companies and large retailers, first emerged in the
1950s, but in most instances full payment were expected within the billing cycle. In the 1960s,
retailers began converting their charge cards into credit cards, a credit instrument that allowed the
consumer to extend payments over a long period. Generic revolving credit, that is, a re-usable credit
account not tied to a specific retailer, dates to the 1950s with the emergence of the first bankcards,
but it begin to flourish with the introduction of credit cards carrying the Visa and MasterCard logos;
its usage more than doubled over the 1970s, with much of that growth taking the place of small
installment loans The substitution accelerated in the 1980s and 1990s as credit cards--some tied to
home equity lines of credit--became widely accepted for the purchase of larger durable goods and as
a ready source of funds through cash advance features.
2. Credit information system, CISs
CISs (CISs) compile databases that potential lenders can access to help them evaluating a consumer's
credit application. They provide information to potential lenders about an applicant's credit record,
producing a “credit report” that contains details of the payment and credit history of an individual,
financial accounts and the way they have been managed, as well as other information of interest to
the credit industry12. Reports of CISs help banks stem out misconducts in the banking sector since
customers whose credit reports indicate as having been involved in mal practices are subjected to
stringent terms and conditions. This is also expected to help banks suppress the levels of NonPerforming Loans while increasing their loan books. Credit information sharing to bank customers,
is expected to minimize the problem of information asymmetry in the financial sector. Information
asymmetry between banks and borrowers is one of the main contributors to high cost of credit. Thus
banks tend to load a risk premium to borrowers because of lack of customer information. This in turn,
increases cost of borrowing, meaning repayment of loans escalate which translates to a high level of
default.
12
Ferretti, (2006)
191
The Credit Information Sharing (CIS) apparatus is therefore expected to facilitate the development of
information capital to reduce information asymmetry or increase information symmetry and allow
cost of credit to decline substantially. It is therefore the Central Bank’s expectation that savings arising
from the sharing of credit information will translate to lower cost of credit. CISs assist lenders to make
faster and more accurate credit decisions. They collect, manage and disseminate customer information
to lenders in the form of credit reports. These credit reports will help lenders to decide whether to
extend an applicant’s loan, credit card overdraft facility or extend any other product, which is reliant
on customer’s ability to repay at a determined cost.
3. CISs and Commercial Banks
In establishing the CISs, what needed to be done first was to convince banks and other financial
institutions that if one institution benefits, they all benefit13. Customers are then well served and,
consequently, receive products that they can afford. Thus there will be fewer loan losses, as the credit
institutions loan money responsibly, and then fewer write-offs. In the end, much as with the fraud
detection models, savings can be passed on to customers in the form of lower interest rates and better
customer service. However,14 cautions that although individual banks may find it hard to resist
following these trends as a result of market pressure, such an increased homogeneity of business
models may augment the vulnerability of the banking sector as a whole.
The individual financial institutions can use the information from the CISs for credit scoring and
evaluating client credit worthiness. The process of modeling the variables important in the extension
of credit is referred to as credit scoring 15.Based on statistical analysis of historical data of the
customers; certain financial variables are determined to be important in the evaluation process of a
credit applicant’s financial stability and strength. This analysis produces coefficients which are
translated into score weights. Subsequently, information on these important variables is obtained for
new bank customers. An overall score for these new applicants is produced by adding the weighted
scores which were generated from the responses to the different variables. If this overall score is above
a predetermined cut-off point, the loan applicant receives a certain line of credit. If not, the applicant
is denied credit.
Commercial Banks in Albania are financial institutions that are authorized by law to receive money
from businesses and individuals and lend money to them. They are open to the public and serve
individuals, institutions and businesses. They are mainly established with the aim to make a profit by
carrying out these activities. Their operations are licensed, supervised and regulated by the central
bank.
4. CISs based on Credit Scoring Models produce credit scores
A credit score is a numerical expression based on a level analysis of a person's credit files, to represent
the creditworthiness of the person. A credit score is primarily based on credit report information
typically sourced from CISs. Lenders use credit scores to evaluate the potential risk posed by lending
money to consumers and to mitigate losses due to bad debt. Lenders use credit scores to determine
who qualifies for a loan, at what interest rate, and what credit limits. Lenders also use credit scores to
determine which customers are likely to bring in the most revenue. The use of credit or identity scoring
prior to authorizing access or granting credit is an implementation of a trusted system. Credit scoring
can be formally defined as a statistical (or quantitative) method that is used to predict the probability
that a loan applicant or existing borrower will default or become delinquent16. This helps to determine
whether credit should be granted to a borrower 17.Credit scoring can also be defined as a systematic
13 Leonard, (1996)
14 Cavelaars and Passenier,(2012)
15 Leonard, (1995)
16 Mester,1997
17 Morrison, 2004
192
method for evaluating credit risk that provides a consistent analysis of the factors that have been
determined to cause or affect the level of risk 18.The objective of credit scoring is to help credit
providers quantify and manage the financial risk involved in providing credit so that they can make
better lending decisions quickly and more objectively. In the United States, the Circuit Court has
found considerable actuarial evidence that credit scores are a good predictor of risk of loss 19.
Similarly, a recent actuarial study has concluded that credit scores are one of the most powerful
predictors of risk; they are also the most accurate predictor of loss seen in a long time 20.
2.2 CISs impact borrower repayment and reduce NPLs
According to 21the private sector credit relative to GDP is positively correlated with information
sharing in their study of credit market performance and institutional arrangements in 129 countries
for the period 1978–2003. Firm-level data suggest that information sharing may indeed have a
differential impact on credit availability for different firm types. 22Combine cross – sectional firm level data from the 1999 World Business Environment Survey with aggregate data on private and
public registries collected in [Miller (2003)]. They find that private credit bureaus are associated with
lower perceived financing constraints and a higher share of bank financing (while public credit
registries are not), and that these correlations are particularly strong for small and young firms.
To remain competitive, CIS worldwide must not stand on their laurels; they must introduce innovative
services to meet the evolving needs of their clients. The impact of credit rating or scoring agencies on
financial markets has become one of the most important policy concerns facing the international
financial architecture. Ratings indicate a relative credit risk and serve as an important metric by which
many investors and regulations measure credit risk.
23Find empirical evidence that the lending market would collapse due to credit risk in the absence of
information sharing institution and reputational banking. However, their study also showed that
establishing CISs encouraged borrowers to repay their loans by allowing lenders to identify borrowers
with a good payment history. Their study showed that an information sharing institution positively
impacted the credit market in the following ways: Without CISs, borrowers had a tendency to repay
loans only when they planned to maintain their current lending relationship. However, in economies
with a credit information institution, borrowers had a higher chance of repaying their loans
regardless of whether they were planning to continue their current lending relationship or not. Thus,
it can be implied that credit sharing institutions, by documenting borrower behavior, can positively
impact borrower repayment and reduce NPLs.
The Credit Register in Albania became effective in February 2008. The Regulations require all
licensed banks and non-bank financial institutions to share information on Non-Performing Loans
(NPLs) through the Credit Register which is part of the Central Bank of Albania (CBA). The role of
Credit Register is to collect, collate and process data received from approved sources of information
and generate credit reports to be used by lenders.
Conclusions
CISs services assist in reducing the incidence of non-performing loans. This is made possible through
the factors of transaction costs, competitive information sharing, loan loss and delinquency, credit
evaluation practices that are enhanced when the CIS services are used.
The implementation of CISs reduces transaction costs and certainly borrowing costs for consumers.
Lending institutions in Albania need the CISs services as they enhance their profitability by reducing
transaction costs involved in identifying suitable clients that the bank can give loans.
18 Fensterstock,2005
19 Johnson-Speck, 2005
20Miller, 2003.
21 Djankov,(2007)
22 Love and Mylenko (2003)
23 Brown and Zehnder (2007)
193
Limitations of Credit scoring model should also be noted. One of the major problems that can arise
when constructing a credit scoring model is that the model may be built using a biased sample of
consumers and customers who have been granted credit 24. This may occur because applicants (i.e,
potential customers) who are rejected will not be included in the data for constructing the model.
Hence, the sample will be biased (i.e, different from the general population) as good customers are
too heavily represented. The credit scoring model built using this sample may not perform well on
the entire population since the data used to build the model is different from the data that the model
will be applied to. The second problem that can arise when constructing credit scoring models is the
change of patterns over time. The key assumption for any predictive modeling is that the past can
predict the future25.In credit scoring, this means that the characteristics of past applicants who are
subsequently classified as “good” or “bad” creditors can be used to predict the credit status of new
applicants. Sometimes, the tendency for the distribution of the characteristics to change over time is
so fast that it requires constant refreshing of the credit scoring model to stay relevant.
One of the consequences of credit scoring is the possibility that end-users become so reliant on the
technology that they reduce the need for prudent judgment and the need to exercise their knowledge
on special cases. In other instances, end-users may unintentionally apply more resources than
necessary to work the entire portfolio. This could run into the risk of a self-fulfilling prophecy26In the
U.S., a new industry has emerged that is dedicated to help borrowers improve their credit scores by
rearranging finances27 rather than obeying the simple rule: pay your bills on time and keep your debt
low. Such score-polishing actions could potentially distort the patterns of credit default. Despite the
limitations of Credit Scoring models highlighted above, there is no doubt that credit scoring will
continue to be a major tool in predicting credit risk in consumer lending. It is envisaged that
organizations using credit scoring appropriately will gain important strategic advantage and
competitive edge over their rivals. From a management perspective, it is worthwhile noting that credit
scoring has the following benefits: (1) reduced subjectivity and increased objectivity in risk
assessment, (2) increased speed and consistency of risk assessment, (3) possible automation of the
risk assessment process, (4) reduced cost of risk assessment, (5) better allocation of resources, (6)
better determination of interest rate, loan duration, loan amount, etc., (7) better risk management, (8)
better collection/recovery strategies, and (9) possible development of non-traditional markets. It is
hoped that this paper can make a contribution to the credit scoring literature.
Data mining has gained widespread attention and increasing popularity in the commercial world.
Besides credit scoring, there are other potential data mining applications for businesses. For example,
data mining can be used to: (1) perform churn modeling to identify customers who are likely to churn,
(2) construct fraud detection models to give early warning signals of possible fraudulent transactions,
(3) understand consumers and customers better (e.g., via market basket analysis), (4) segment
customers (e.g., via clustering), or (5) construct models to predict the probability of purchasing certain
products or services in order to facilitate cross-selling or up-selling. The findings can then be used,
say, to prepare mail catalogs, advertisements, promotion campaigns, etc.
Policy Recommendations
1. The implementation of CISS reduces transaction costs and certainly borrowing costs ultimately the
government needs to craft strategies and policies to build CIS in Albania and to oblige relevant
institutions to report the consumers data into CIS and lending institutions to get CIS services as this
will benefit the populace through reduced interest rates, banks will also benefit from reduced
information asymmetry and nonperforming loans.
2. To enhance credit information sharing, information access should be available at low or no cost
and this would be facilitated by creating an environment that supports more competitive information
sharing both financial and non-financial institutions should be allowed access to credit information
reports of borrowers.
24 Hand, 2002
25 Berry and Linoff,2000
26 Lucas, 2002
27 Timmons, 2002
194
3. Central bank should regulate the CISs to give up to date credit information about borrowers which
may increase their effectiveness and eventually would reduce information asymmetry, costs and
translate into reduced adverse selection.
References
1. AkSIKe, H. [1973]: Statistical predictor identification, in: Ann. Inst. Statist, Math. 22, 203-217.
2. Altman, E.I. [1968]: Financial Ratios, Discriminate Analysis and the Prediction of Corporate
Bankruptcy, in: The Journal of Finance 23 (4), 589-609.
3. Comerce banks in Albania (annual bulletin)
4. A Handbook for Developing Credit score Systems in a Microfinance Context. Prepared by DAI
Washington for USAID. October 2006.
5..http://www.microlinks.org/ev_en.php?ID=13456_201&ID2=DO_TOPIC
6. Chawanda, Rob (OIBM) and Angenita de Rijke (ING).Opportunity International Bank of Malawi:
First Step in Credit score. October 2007.
7. http://www.oinetwork.org/documents/First%20Step%20in%20Credit%20Scoring.doc
8. Mark Schreiner (Microfinance Risk Management) and Dellien, Hans (Women’s World
Banking).“Credit score, Banks, and Microfinance: Balancing High-Tech With High-Touch.‖
December 18, 2005.
9. Canova, L..(2007). the many dimensions of poverty in Albania: Income, wealth and perceptions.
10. EUruçi, I. Gedeshi. (2003). Remittances management in Albania . (2003). CESPI, Center for
International Policy Studies. CESPI,
195
Performance Indicators Analysis inside a Call Center Using
a Simulation Program
Ditila Ekmekçiu1, Markela Muça2, Adrian Naço3
1,2
University of Tirana, Faculty of Natural Sciences, Albania
3
Faculty of Engineering Mathematics and Physics
ditila.ekmekciu@gmail.com1, markela.moutsa@yahoo.com2
Abstract. This paper deals with and shows the results of different performance indicators analyses
made utilizing the help of Simulation and concentrated on dimensioning problems of handling calls
capacity in a call center. The goal is to measure the reactivity of the call center’s performance to
potential changes of critical variables. The literature related to the employment of this kind of
instrument in call centers is reviewed, and the method that this problem is treated momentarily is
precisely described. The technique used to obtain this paper’s goal implicated a simulation model
using Arena Contact Center software that worked as a key case at the time where the events analyses
could be executed. This article comes to the conclusion that Simulation is a completely suitable
instrument to accomplish its purpose since it could be adequate to demonstrate, for the call center
taken into consideration principally that: (a) it is feasible to reduce the agent contingent; (b) reasonable
variations on the demand design can impact very much the key performance indicators of the Call
Centers’ project; and (c) it is feasible to increase the service level if a combined handling format is
followed.
Keywords: Call center, Simulation, Queueing Models
1. Introduction
Call centers are operational centers created with the purpose of using both Communication and IT
resources, in order to make automatic very big volumes of different activities and telephone services,
not only of incoming calls, but also those generated by the center as well (outbound activities). The
inbound call centers, where we have incoming calls (so are the clients who call the call center), are
characterized by a system compounded by many call attendants that receive calls from other persons,
usually clients of that service, or even potential clients that wish to get information regarding the
specific subject, buy a specific product, look for technical assistance, answer to a certain research,
update known data, register incidents or place a complaint, between other requests (GROSSMAN et
al., 2001 [1]; HAWKINS et al., 2001 [2]).
As reported by Mehrotra, Profozich and Bapat (1997) [3], managers of call centers have a much more
difficult job today than they used to have in the past. With numerous products and services being
particularly created, made available and ready to use, sold and supported by technicians out in the
market, the call centers have struggled to provide various service levels for different types of clients,
which have different needs. At the moment, the telephone systems allow a high flexibility and can
decide how the calls can be routed and put on line. However, at the same time, this makes the planning
and analyses even more sophisticated because of the fact that they make possible a link among
multiple call centers, prioritization of certain specific calls, existence of various abilities between
agents and customization of calls routing.
Today managers must be able to know completely what is happening at call centers so they can know
how calls, routes, priorities, agents and their capacities, peak periods and other aspects that affect the
service level and the utilization rates (BOUZADA, 2006 [4]).
Labor costs represent almost 70% of the total costs of the industry, justifying the requirement for an
efficient management and the great importance of a quantitative access for the dimensioning of a
service handling capacity that consists on a trade-off between this cost and the establishing of the
196
right service level; Saying it differently, in having the right number of qualified people and resources
at the right moment, in order to work with the forecast working load, keeping the quality arrangements
and the required service level. This way the use of better accurate models on the dimensioning of the
size of the staff, of the industry that works with big financial volumes, is considered as more relevant
than ever (HALL; ANTON, 1998 [5]).
As studied by Mehrotra and Fama (2003) [6], and Hall and Anton (1998) [5], call centers are
interesting objects for simulation studies, because: (a) they are faced with more than one type of call,
where every type represents a line; (b) the calls received in each line arrive at random, as time passes;
(c) in some cases, agents make calls (particularly in telemarketing), or as a return for a call received;
(d) the duration of every call is random, as the after call work of the agent ( data collection,
documentation etc.); (e) the evolution on the systems that route the calls for the agents make the
philosophy behind the call center even more sophisticated; (f) agents can be disciplined to answer
only one kind of call, several kind of calls or all kinds of calls with different priorities and/or
preferences described for the routing logics; and (g) the high amount of money invested in call centers
on capital and work, is able to justify the application of this strong tool.
According to Hall and Anton (1998) [5], call centers can use Simulation to test (and eventually
legitimize its implementation), if particular alterations can improve the system before its
implementation. The most important call centers use this instrument effectively and efficiently,
making possible to project the system, to manage the operation and to plan the future, regardless of
possible scenarios.
Based to this fact, this paper describes the dimensioning problem of the handling capacity of a large
Albanian call center, looking for an immediate proposal presentation of different scenario analysis,
altering some key parameters of the system. Such analyses are made possible by the use of Simulation
during work, whose goal is to calculate the sensitivity of the call center’s performance to eventual
changes of critical variables.
2. Literature Review
In accordance to Anton (2005) [7], the main cost in a typical call center is due to human resources
(almost 70% of the total costs), far above those regarding the technology (16%), the second in the
expensive classification.
For this reason, and in order to reduce staff requirements, one of the most important duties consists in
managing the call centers queues that occur when there is no agent available to handle a client, who
waits on a virtual line from which he will leave only when an agent is free to attend him or when he
disconnects the call. As showed by Brown et al. (2002) [8], in the case of call centers, the virtual
queue is invisible between the clients and between the clients and the agents.
In the call centers scenario, Araujo et al. (2004) [9] say that the queues discipline, when well managed, is an important partner of the call centers production workforce management area, which
have as a goal to achieve the wonted results with finite resources, turning this area very important for
these companies. In accession, a significant reduction in the clients waiting time can be achieved.
Some of the call center’s characteristics make it difficult to apply analytical formulas of the Queuing
Theory for its modeling, including: generic distribution for the handling time, time-varying arrival
rates, temporary floods and calls abandonment (CHASSIOTI; WORTHINGTON, 2004 [10]).
Chokshi (1999) [11], Klungle and Maluchnik (1997) [12], Hall and Anton (1998) [5], Mehrotra and
Fama (2003) [6], Avramidis and L’Ecuyer (2005) [13], Klungle (1999) [14] and Bapat and Pruitte Jr.
(1998) [15] go further a few current factors that contributed in the increase on the demand for the use
of the Simulation tool in the call centers sector: (a) the growing importance of the call centers for a
big number of corporations, due to the fast increase of information, communication and technological
devices, increasing the need of using scientific methods in decision makings and instruments for its
strategic management rather than using the intuition, only; (b) the increasing complexity of the traffic
of the calls ahead with rules more viewed on the skill-based routing; (c) the uncertainty that is more
predominant in the decision problems commonly found in the operational management of call centers
phone desks; (d) fast changes in the operations of the company and the improvement in the reengineering projects resulting from the growth of mixed companies and of acquisitions, business
unpredictability, outsourcing options and the utilization of different channels in order to reach the
197
costumer (telephone, e-mail, chat etc.); and (e) the availability and accessible price of the computers
available in an everyday market less complicated, intuitive and easier to be absorbed and used.
The Simulation, according to Mehrotra (1997) [16], certainly shapes the interaction among calls,
routes and agents, the random individual incoming calls and the random duration of the handling
service, too. Through the use of Simulation, managers and analysts interpret the call centers gross
data (call forecast, distribution of the handling times, schedule hours and the agents skills, call route
vectors, etc.), in handling information on service levels, clients call abandonment, use of agents, costs
and other valuable performance measures of a call center.
In accordance to Chokshi (1999) [11] and Klungle and Maluchnik (1997) [12], the use of Simulation
to help management in decisions making in a call center permits the coming benefits: (a) visualize
future processes and use it as a communication tool; (b) validate the processes premises before its
implementation; (c) analyze the impact of the changes in details; (d) predict the aggregated needs of
resources and schedule the working staff; (e) measure the performance KPIs; and (f) estimate the
impacts on costs and economies.
The first use of the Simulation in a call center, as mentioned by Hall and Anton (1998) [5], is the
evaluation when one can confirm “where the call center is”. The basic argument is “how efficient is
the operation today?” The purpose of this evaluation is to organize a point of start (and reference) for
the change.
Mehrotra, Profozich and Bapat (1997) [3] and Yonamine (2006) [17] discuss about the other utility
of Simulation in a call center: the Simulation permits fast and accurate understanding of how the
operational performance of the call center would work when confronting specific scenarios (based on
modifications caused by several initiatives such as the adoption of a new technology, a new business
strategy or the increase of the volumes of work), before any change is truly made, but not intervening
on the operation of the call center’s phone workstations and not impacting its budget, too. This way,
a few questions might be answered among others: (a) which is the impact of a call overflow? (b)
Which are the compromising in the act of prioritizing special clients? (c) Will the service improve if
agents provide main pieces of information to clients? (d) Which are the potential achievements related
to the adoption of a predictor dial?
According to these authors and to Gulati and Malcolm (2001) [18], Bapat and Pruitte Jr. (1998) [15]
and PARA¬GON (2005) [19], a simulation model can be used and has been used more commonly
than ever to plan a few other critical aspects of the modern call centers of all sizes and types, such as:
(a) a particular service level; (b) flexibility on the time distribution between incoming calls and of
handling time; (c) consolidation of the central offices; (d) skill-based routing; (e) different types of
calls; (f) simultaneous lines; (g) call disconnect patterns; (h) call returns; (i) overflow and filling of
capacity; (j) waiting lines prioritization; (k) call transference and teleconferences; (l) agents
preferences, proficiency, time learning and schedule. The outputs model can emerge in format of
waiting time, call disconnecting average amount, (both with the possibility of distinction on the call
types) and level of the agents employment (with possibility of the agent types distinction). And, due
to the application of this approach to the real and complicated characteristics of call centers, the
Simulation can make its dimensioning and management more trustworthy.
Mehrotra and Fama (2003) [6] and Klungle (1999) [14] imagined future propensities capable to
impact the simulation of call centers, such as: (a) the operational complexity, which will increase
continuously – more waiting lines, more variation on the agents scale and combination diversity
among skills and route rules – forcing the analysts to create more powerful models; (b) emerging of
more Simulation software’s specialized in call centers, whose importance tends to follow the role that
the Simulation will assume in the process of remodeling the central offices that are necessary to the
dealing with the new complexities; and (c) a greater understanding by the executive managers that the
call centers are main components of the clients’ value chain, discharging a wish to understand the
main risks of any operational configuration and the resultant improvement of the quality of the
collected data and precision of the parameters (such as distribution of time between incoming calls,
handling time, waiting time, average of disconnecting etc.), holding more healthy results.
Gulati and Malcom (2001) [18] used Simulation to compare the performance for three different calls
programming approaches (heuristic, daily batches optimization and dynamic optimization),
displaying opportunities for the improvement of the outbound call center process within the studied
bank. The model outputs gave a way to check the system’s performance compared to the management
objectives and showed that the non-heuristic approaches obtained better results, but not during the
whole day.
198
Miller and Bapat (1999) [20] described how Simulation was used to project the Return Of Interest
related to the acquisition and utilization of a new call routing technology for 25 call centers.
Requesting US$ 17 million of investments and an operation cost of US$ 8 million per year, it was
demanded to check if the technology would cause enough benefits (cost reduction, agents productivity
increase and possibility to handle more calls) in order to assure its implementation on a national range.
Lam and Lau (2004) [21] wrote about a restructuring effort of a Hong Kong company which supplies
service for computer and office equipment. As long as there were many opportunities accessible to
improve the process, Simulation was used to explore the different options and judge the results of the
existing call centers restructuring. The simulated results analysis confirmed that the great improve
opportunity consisted of the juncture of the current resources in a singular call center.
Saltzman and Mehrotra (2001) [22] showed a study where Simulation was used by a software
company which meant to visualize its call center operating before the launching of a new paid support
service program. They wanted to verify if the objective– the paying customers waiting less than 1
minute before being handled – would be achieved. The management also wondered which would be
the new program impact to the service given to the non-paying customers’ regular basis.
3. The Case
3.1 The company
The Call Center taken into consideration is Teleperformance Albana, part of the corporate
Teleperformance, a leader in this sector in the world. It is present in 62 countries, with 270 call centers,
created 36 years ago, that invoiced in 2014 $3.7 billion. Teleperformance Albania was born in 2008
as part of Teleperformance Italy, operates in Tirana and Durres and has 2000 employees. It was rated
the sixth company with the highest number of employees in the country. Being part of the Italian
company, it operates in the Italian market with very important clients. The dimensioning process of
handling capacity The dimensioning consists in the analysis that may customize physical, technical
and staff structures of a call center against the objectives of the customer service operation that begins
with the forecast of the demand inside the days. The Lottery campaign was chosen to demonstrate the
dimensioning problem, since its demand is the most predictable and, as a result, being possible to
measure the quality of a dimensioning process independently i.e., starting from the assumption that
the input – demand forecast – introduces a good quality. In addition to that, there are only two types
of clients of the Lottery campaign: extra (a paid service) and main (a service free of charge). The
service level for this project is related to the waiting time of the final client in line, from the moment
the incoming call arrives in the system to when it is answered from the agent. Saying it differently, it
is the time which the client remains in line, listening to the background song and waiting for the agent.
More specifically, the service level consists in the percentage of calls that wait no more than 10
seconds to be answered. Since only the calls answered count in this computation of the service level,
the disconnections are not taken into consideration (and as a consequence not punished), for effects
of the service level.
However, they are measured through another indicator that is the abandonment rate and our Call
Center pays penalties when this rate surpasses 2% in a month. As this can happen, to avoid the
disconnection is considered as a priority, to the damage of the service level, as long as it is kept above
a minimum value. The service level does not include legal requirements of the contract (like the
abandonment rate), but does affect the commercial relationship; i.e., it is interesting to not give
priority only to the abandonment and, as a result, not consider the maintenance of the service level in
decent values.
The dimensioning routine – isolated for each product (main and extra − due to the priority of the last
over the first) – starts with the computation of the daily needs of the agents, departing from the
forecasted calls, the average handling time (AHT) and the average time during which the agents are
busy per day. After that, the need of the agents (adapted to the 6-hours-agents pattern) is compared to
the resources availability, discounting the losses regarding absenteeism (vacations, sicknesses or not
justified absences). The result of this comparison is the balance or the deficit of the work for each day
of the projected month. The output of this first step is the amount of agents that have to be hired or
discharged in the indicated month so that the required numbers can be obtained.
199
From the moment the contract decision is taken, or the discharge is decided and implemented, the
planning staff can go through a more detailed analysis – the daily dimensioning. This must be done
for a day only, and this designed format should be repeated for the other days of the considered period,
as long as the scheduled hours of each agent should be the same every single day of the specific
month.
Concluding, a volume of calls and an average handling time (necessary numbers for the dimensioning)
should be chosen to be used as a diagram for the dimensioning of all days of the month. The chosen
day for the diagram is, generally, the fifth day of higher movement. Acting like this, the dimensioning
will guarantee the desired service level for this day and all the other days with lower movement, but
not for the four days of higher demand, when there will be a loss in the service level. Although, this
doesn’t introduce a problem, because the agreement related to the Lottery includes a monthly service
level and not a daily service level.
For the day chosen as a pattern for the dimensioning of the month we applied a curve that should
indicate the daily demanding profile, i.e., what daily volume percentage will happen during the first
half hour of the day, during the second half hour of the day, …, and during the last half hour of the
day. This kind of curve is shown based on the calls report taken at each period of half hour for each
day of the week. Regarding the Lottery, the curves of every day of the week are very much alike
(principally from Monday to Wednesday, with a little increase of the volumes during afternoon of
Thursdays and Fridays), and during Saturdays and Sundays they are a little different.
The result of this operation is a forecast call demand (volume and AHT for each half hour). Using the
concepts of the Queuing Theory and with the support of the Excel Supplement Erlang formulas, called
Turbo Tab, it is calculated the necessary amount of agents who will be handling the demand of each
period with a minimum pre-settled service level (usually 85% of the calls being answered before 10
seconds).
The last month contingent of agents is then taken into consideration. Because of the amount of agents
that are starting their work at each day period and the daily work load of each one of them (4 or 6
hours), a sheet calculates how many agents will be available for each period of half hour. This
information is then compared to the agents’ requirement for each period of 30 minutes, formerly
calculated. Over the actual agents scale, the planning team will work on modifying the agents’
availability for each period of the day, in order to obtain the desired service level. The goal is to
persuade a specific amount of people in each scheduled hour, during a trial-and-error process, over
which it will be necessary to analyze several factors, like daily working hours load, working laws
aspects and available workstations. In the case of the Lottery, the balanced scale (varying times with
the operational staff over or under the requirements) can be utilized, since what really matters for
commercial scopes is the daily average level service. Throughout the staffing process, the planning
department makes experiments by changing the quantity of agents that start working at each period
of time. These changes therefore modify the quantity of agents available in each half hour period. The
sheet containing the Erlang formulas uses then this information to predict the service level for each
half hour period and for the whole day that depends on the forecast demand, too.
During this interactive process, the principal motivation of the analyst is to maximize the day’s
average service level. The service level during each hour band, itself, doesn’t present a big worry to
the analyst who, however, tries to avoid great deficits of agents assigned in comparison to the
demanded within the hour bands of the day. The worry about daily deficits exists because, during the
hours with a higher deficiency of agents it is possible to register a great occurrence of abandonments.
And of course this can be very bad for two main reasons: penalties for surpassing the call
abandonments and the possible fact that the client that didn’t get an answer returns the call later on
and waits until getting an answer, as a result degenerating the service level. This dimensioning effort
main objective is to provide a better adjustment between the demanded and offered capacity during
the day. During the last part of the dimensioning and staffing processes, the analyst tries to estimate
how the operation service level will be (percentage of calls answered in less than 10 seconds), on all
days of the month (until here the calculation was based on the fifth day of higher movement, only).
The distribution within the day of the agents elaborated during the past steps is repeated on all days
of that month and, ahead with the daily call demanding forecast as well as with the demand within
the day behavior profile, is capable, as a result, to estimate − through the Erlang Methodology − the
service levels to be achieved for each day and hour, within the specific month.
Methodology applied to perform the analyses (scenario and sensitivity)
200
For the real world of call centers, the Queuing Theory is the best analytical methodology to be applied,
but there are experimental methods – such as Simulation, for example − that should be even more
satisfactory for an industry with an operational day to day as complicated as modern call centers, as
recommended by section 2 of this paper.
The utilization of the Simulation permits us to consider the displayed characteristics of the same
section, including the abandonment behavior (it is possible to consider that a percentage of clients
who disconnected their calls, will return and try a new contact within a given quantity of time that
can be modeled using a statistical distribution) and a flexibility on the definition of the handling time
distribution.The concept consists in simulating by computer and in a little time, the call center’s
operation work during periods of 30 minutes. Acting like this, it is not necessary to experience in
practice some of the dimensioning alternatives so that we can know the consequences because the
experimentation is made in a virtual and not physical environment. However, it is possible to see the
operation (with the calls arriving, being sent to the queues and then handled) and what would happen,
in detailed forms, so that to understand why a specific period of the day presented a service level so
low/high, for example (instead of only accepting the number provided by the analytical formulas).
For the dimensioning and staffing of the agents to handle the extra clients of the Lottery, in September
2015 it was utilized the assumption (originated on the demand forecast) that 586 calls would come to
the phone workstation with an AHT of 29 seconds in the first half hour of the day (from 00:00 a.m.
to 00:30 a.m.). The staffing team requested then 12 agents for this period.
In the software Arena Contact Center it was built a model to simulate how the system would behave
in this time, with the same demand assumptions (volume and AHT) and with the same operational
capacity (12 agents).As the calls come to the phone workstation without any type of control, this
process can be considered as a random one, the conceptual basis suggesting as a result that the call
arrivals rate could be shaped through a Poisson process. The imagined simulation model implemented
this process with a mean of, approximately, 0.33 calls arriving per second (or 586 in a 30 minute
interval).In regard to the handling time, it is used the Erlang distribution to better shape this process,
and, as a result, it was considered with a mean of 29 seconds. However, it requires an additional
parameter (k) linked to the variance of the data around the mean. The standard deviation of the
distribution is equal to its mean divided by the square root of k. To be capable to consider a moderate
variance of the data around the mean, the model takes the Erlang distribution with k = 4, resulting on
a variation coefficient of 50%. In order to allow a right interpretation of the clients’ abandonment
behavior, it was essential to perform a research close to the Teleperformance Albania basis that
includes the disconnected calls of the Lottery. The research demonstrated that the waiting time of the
calls disconnected historically introduce a mean of about 2.5 minutes, keeping a distribution not very
far from an exponential one. It was also fundamental to model the return behavior of the disconnected
calls. In order to do this, it was used the premise that 80% of the disconnected calls are recalled
between 1 and 9 minutes after the disconnection (a uniform distribution).
The simulation of the call centers’ operation during 30 minutes was replicated 100 times in the
software in a period of 142 seconds, and the first results show that, in average, 595 calls were
generated in each replication. This number is a little bit higher than that demand premise of 586 calls
because in the simulation, some of the disconnected calls were replicated and put into the queue again.
From the generated calls, 579 calls in average were effectively handled by the agents in each
replication, generating an AHT of 29.35 seconds. From these calls, 541 were handled before 10
seconds, giving a service level of 93.31%. From the 595 calls generated in each replication, 14.5 (in
average) were disconnected by the clients, producing an abandonment rate equal to 2.44%. Between
the disconnected 14.5 calls, 11.5 (79.41%) returned to the queue a few minutes after the disconnection.
In average, the agents were busy 78.75% of the time, during this period.
From this key scenario, different scenario and sensitivity analyses were studied for a better
comprehension of the system operational behavior dealing with potential changes of its main
parameters. This methodological experiment is very much alike to the ones used and described by
Miller and Bapat (1999) [20], Gulati and Malcom (2001) [18], Saltzman and Mehrotra (2001) [22],
Lam and Lau (2004) [21] and Yonamine (2006) [17], whose works were indicated during section 2
of this paper.
Analysis of scenarios, sensitivity and results
201
All previous results start from the premise that the handling time follows an Erlang distribution (with
a variation coefficient equals to 50%). However, it is possible to have the handling time demonstrating
a different variance and that this parameter can cause an impact on the most important results.
The sensitivity analysis regarding the variance of the handling time tries to measure this impact. The
same simulation was replicated a few times using the software, every time with the same mean on the
handling time (29 seconds), but with different values for k (and, as a result, for the variation
coefficient), from where the appropriate outputs were collected, and the principal performance
indicators (service level and abandonment rate) could be achieved, which are shown on the following
Table 1.
As k increases, the variance of the handling times reduces. Due to the uniformity of these times, the
system becomes more stable, presenting as the most evident consequence an increase of the service
level. But the abandonment rate has not a clear tendency, even though it can look like falling as the
variance of the handling time decreases (k increases). The variation on these outputs is not very large,
but is far from being insignificant, demonstrating a significant potential impact of this parameter on
the most important results. However, the right consideration of the handling time variance − and not
only of its mean – shows to be extremely necessary in order to achieve accurate results.
Table 5. Service level and abandonment rate for various values for the parameter k of the Erlang
distribution, from 00:00 a.m. to 00:30 a.m., Sept/15, 12 agents
Source: Table elaborated from the results achieved by the software.
The worst performance (service level = 91.03% and abandonment rate = 2.71%) happened in a
situation in which the variance was the largest possible (k = 1; variation coefficient = 100%). This is
an Erlang distribution case that corresponds with the exponential distribution, the same format used
to model the handling time in the analytical methodology employed by Teleperformance Albania to
estimate the indicators.
This evidence awakens a curiosity related to the verification of the results of a simulation that takes
into account another kind of distribution − since, according to what was indicated on section 2, the
behavior of the handling time in call centers can show different formats − generally used, as well, to
model this variable: the lognormal. Adopting the same original simulation model, but changing this
variable distribution to fit the highlighted format (with the same mean – 29 seconds, as well as
maintaining the same variation coefficient of 50%), 100 replications were run in the Arena Contact
Center software.
In average, 597 calls were generated, 583 handled (544 before 10 seconds) and 13.95 abandoned. The
resulting service level and abandonment rate were respectively 93.40% and 2.34%. These indicators
are very close to those achieved with the Erlang distribution where k = 4 (respectively 93.31% e
2.44%), giving a certain accuracy to these values and suggesting that any of the two formats frequently
used to model the handling time can be used without making a distinction.
Simulation also permits the scenario analysis (What-if). At the shown example being studied here,
since the employment of 12 agents in the highlighted period generated a service level (93.31%)
adequately higher than the minimum objective (85%), what would happen with this indicator after a
reduction of 1 agent? Would it be possible to maintain it above the objective?
Using this scenario, an average of 576 calls was handled, from which 491 before 10 seconds. The
service level achieved in this scenario with 11 agents was then 85.23%. This value continues to be
above the 85% settled for the extra clients. In other words, the 12th agent was not so necessary (in
terms of achieving the service level objective), despite his absence lowered the service level in more
than 8 percentage points.
But it wouldn’t be bad to also know the impact caused by this kind of reduction in the abandonment
rate. And, from the 604 calls generated in each replication, 26.2 – in average – were abandoned by
202
the clients, resulting in rate equal to 4.34%, revealing a high impact on this performance indicator.
However, if this indicator was not to be treated as so important, the utilization of Simulation could
cause the savings of one agent for this time band that can effectively occur in some scenario in which
the abandonment rate can be seen in a lower level.
The agents’ reduction impact can also be seen in case of greater deficits of handling availability
through a more complete analysis of sensitivity. The same simulation was repeated a few times in the
software, always with the same parameters, but changing the number of agents. The important outputs
were collected, and from them the principal performance indicators (service level and abandonment
rate) could be achieved, which are shown on the following Table 2.
Table 2. Service level and abandonment rate for different amounts of agents, from 00:00 a.m. to
00:30 a.m., Sept/15
Source: Table elaborated from the results achieved by the software
As it was expected, when the handling availability decreases, the service gets worse, its performance
indicators, too, mainly the abandonment rate that quadruplicates after a 2 agents reduction. With this
contingent, the service level is much lower than the objective, but still cannot be considered
inadmissible, which occurs in the 9 agents scenario, when it becomes 3 times lower than the original
value. With this handling availability, the amount of abandoned calls surpasses those ones handled
before 10 seconds, making the abandonment rate almost as high as the service level!
That shows a great impact of the reduction of the amount of agents on the system performance
(revealing the need of the dimensioning activity to be developed with much care); and suggests that
hiring 10 agents for this time band would be the fundamental minimum, characterizing a scenario for
which the performance indicators would be bad, but not destructive.
The most delicate decision on how many agents to hire effectively for this time band should take in
account the potential costs affected in the inclusion/exclusion of 1 or more agent/s on/from the map
of scheduled times. This way, Teleperformance Albania could question if it would be willing to spend
one additional monthly labor cost in order to improve the service level and the abandonment rate for
the highlighted time band by the amounts indicated in the analysis.
As described in section 3.2, the dimensioning of the operational capacity is made individually for
each product – main and extra. However, the Simulation allows – in a very much similar way to what
described Saltzman and Mehrotra (2001) [22] – a judgment of what would happen with the operation
in a scenario in which different clients − main and extra − could be handled under an accumulated
form (by the main and extra agents), but keeping the priority for the extra clients and, proceeding this
way, breaking up the queue discipline normally used by analytical models – “First In First Out”.
Regarding the sizing and staffing of agents to handle main clients of the Lottery during September
2015, it was used the premise that 399 calls of these clients would come to the phone workstations,
with an AHT of 34 seconds on the first half hour of the day (from 00:00 a.m. to 00:30 a.m.), for which
was determined a staff of 7 agents.
The idea now reposes on simulating − in order to examine the system behavior – the scenario where
these calls would be aggregated to the calls of the extra clients during this same time band (whose
premises were described before, in this section), creating a singular queue. The 12 extra agents and
the 7 main ones would be capable to handle both kind of calls, but with different abilities and
priorities, that characterize the skill-based routing, a mechanism feasible only under empirical
approaches, as described before in section 2.
However, main clients do not behave as extra clients do, and their different characteristics must be
planned by the model. Their handling time, for example, is, in average, a little bit higher. For model
scopes, the same Erlang distribution was used for this time, with a variation coefficient of 50%.
Usually, the main client is more patient before disconnecting the call and this waiting time was
modeled as an exponential distribution, too, but with a higher mean of 3.5 minutes (vs. 2.5). It was
203
also considered that a smaller amount of the clients that disconnect the call (70% instead of 80%) try
to recall during a space of time also lower: within 1 and 6 minutes (uniform distribution).
Regarding the priorities, the extra calls are preferential upon main ones and should be handled, as
long as it is possible, by extra agents − theoretically with higher skills − if the extra agents would be
busy at the moment, then the extra calls would be handled by the main agents. The main calls, instead,
would be preferably handled by main agents, theoretically with lower skills, in order to let the best
agents free for more important calls.
In the cases where the calls are not handled by their preferential agents, their handling time is
alternated. The model takes into consideration the fact that an extra call being handled by a main agent
(less capable) lasts 10% more to be handled; on the other hand, if a main call is handled by an extra
agent (more capable), this lasts 5% less to be handled.
The simulation was repeated 100 times in the Arena Contact Center software. In average, 413 main
calls and 591 extra calls were generated in each replication, respectively 19.8 and 9.5 of them being
abandoned. The resulting abandonment rates were 4.80% for the main clients and 1.60% for the extra
clients. Between the disconnected main calls, 14.2 (71.46%) returned to the queue a few minutes later,
and amongst the extra calls, 7.6 (80.44%) did the same. 393 main calls and 581 extra calls – in average
– were effectively handled in each replication. From these, 303 main (77.25%) and 575 extra (98.96%)
were handled before 10 seconds (service level).
Comparing the performance indicators of the extra clients with previous values achieved in the
scenario with segmented handling (abandonment rate = 2.44% and service level = 93.31%), it is easy
to deduce that the aggregated operation became adequately better for these clients. Such like results
were expected, since 7 main agents started handling extra calls. It is real that the 19 agents also
handled main calls, but only when there was no extra call waiting.
This preference permitted a considerable improvement to extra clients handling and, surely, decreased
a little the quality of the service for main clients. But it is interesting to see that, at this scenario, the
service level for ignored clients (77.25%) remained still above the objective to be obtained (75%).
The abandonment rate (4.80%) became a little high, but it is not considered inadmissible for main
clients.
Clear as it looks, the handling aggregated format improved the system performance for extra clients
without ignoring very much the quality of the service for the main ones. This occurred because the
agents could handle their non-preferential calls while ineffective, permitting the increment of the
operation quality as a whole. This type of analyses and conclusions would not be able to be
performed/achieved through analytical methodologies, being possible only by means of an
experimental approach, like the Simulation.
The AHT for main clients was 32.65 seconds, a little bit lower than the value of 34 seconds, used on
the AHT assumption, because a few of them were handled by high-skill agents (extra). For extra
clients, the AHT was of 30.37 seconds, a little bit higher than the 29 seconds assumption, due to the
fact that a few of them were handled by low-skill agents. The main calls waited, in average, 6.24
seconds before being handled, while the extra calls waited only 1.34 seconds. Such difference is due
to the handling preference for the latest calls.
The main agent utilization rate was 89.59%, very close to the extra one (88.84%), mainly because
both kind of agents were qualified to handle both types of calls. Both types of agents were most part
of the time (51-52%) handling the preferential and more numerous extra clients, spending the extra
time answering main clients (37-38%) or ineffective.
The decision on how many agents shall be scheduled and on which assumption they should be
segmented or aggregated in each time band, is completely up to the Teleperformance Albania
planning management. However, it may be probable that the company is concerned on the analysis
of the impact of other variables – that are not under their control (parameters) – on the service level
and abandonment rate for the central desk, which would be possible using Simulation, confirmed in
the section 2.
The scenario analysis can be utilized to find out what would occur with these performance indicators
if − for example − the calls volume during a given time band was 10% higher than the forecast.
Within the simulation of this scenario (but with the same handling staff), 637 calls were handled, in
average, from which 540 – in average – before 10 seconds. The service level for this scenario was
therefore of 84.68%. This value is hardly lower than the 85% established objective for extra clients
and moderately lower than the 93.31%, that would be achieved if the demand had behaved according
to the forecast.
204
From the 672 calls generated in each replication, 32.2 were disconnected, in average, by the clients,
indicating an abandonment rate equal to 4.80%, which demonstrates a great impact of the added
demand to this performance indicator.
So, a simulation of this scenario demonstrated that the original agents contingent (12) − facing an
unexpected 10% demand increase − would be able to practically satisfy the service level objective of
85%, but would also intervene too badly with the abandonment rate.
And what would happen with the service level and the abandonment rate if the demand was
underestimated (also in 10%), even though not linked to its volume, but to the AHT?
In this kind of scenario, 575 calls were, in average, handled; from these, 482 (in average) before 10
seconds. The resulting service level was of 83.85%, a value a little bit lower than the goal (85%) and
fairly lower than the 93.31% that would have been achieved in case the demand had behaved
according to the forecast plan.
From the 606 calls generated in each replication, 29.8 in average were disconnected by the clients. As
a result we have an abandonment rate equal to 4.92%, demonstrating a great impact of the AHT on
this indicator.
Like the scenario that reproduced an increase on the calls amount, the simulation of this situation
revealed that the original contingent of 12 agents − before a suddenly 10% increase on the AHT −
would be capable to ensure a service level almost equal to the 85% objective (in this case, a little more
far), as well as to increase (a little more) the abandonment rate.
Based on the scenario analysis with a deeper demand than the forecast, it is possible to conclude that
a not too large variation (10%) in relation to the forecast values, may impact directly the performance
indicators, particularly the abandonment rate. This conclusion needs much care for the calls amount
and AHT forecast. Another important discovery is related to the probably unexpected fact that the
impact can be even higher when the increase happens with the AHT, when in comparison to a same
magnitude difference on the calls amount. This may lead to a certainty on the AHT forecast being
even more important than the accuracy on the calls volume forecast.
As a result, it could be interesting to explore the impact of higher variations on the AHT (for more
and for less) on the system performance, through a more complete sensitivity analysis. The simulation
of the key model was repeated in the software a few times, but with different values for the mean of
the handling time (from amounts 30% lower to 30% higher), from where the important results were
collected. These results were then organized on Table 3, which follows and also calculates and
presents the main performance indicators, i.e., service level and abandonment rate for each scenario
Table 3. Service level and abandonment rate for different values for AHT, from 00:00 a.m. to
00:30 a.m., Sept/15, 12 agents
Source: Table elaborated with results achieved by software
As a result, higher AHT values than the forecasted ones quickly degenerate the system performance
in terms of service level and abandonment rate (particularly this last one), demonstrating a huge
potential impact of that variable on the most important results. This way, Teleperformance Albania
should dedicate its best efforts to prevent the AHT increase, making their agents awake about the
destructive consequences of an increase on this value.
Analyzing the superior part of Table 3, it is possible to conclude that performance indicators improve
significantly after a reduction of only 10% on the AHT: the service level increases in almost 5
percentage points, surpassing 98%, and the abandonment rate falls to less than its half (1.08%). This
advises that it might be worthwhile to invest on agents training, in order to try to reduce a little the
handling time. Unluckily, it is also true that higher reductions on this time do not hold advantages so
important (particularly for the service level, already found close to its optimum value) for the system.
205
In other words, the cost involved in decreasing the AHT in 10% may be probably compensated by the
benefits resulting from this reduction, but it is difficult to believe that the same would happen with
more huge reductions on this variable.
Similarly to this, there should be other questions forwarding the same type of issues: (a) what would
happen with the abandonment rate if the client became more unwilling to wait and began to disconnect
calls after, for example, 1.5 minutes (instead of 2.5) in average, without being handled? (b) What
would be the impact of this alteration on the service level?
At this new proposed and simulated scenario, 580 calls, in average, were handled, from which 551,
in average, before 10 seconds. The service level for this kind of scenario was, as a result, of 95.09%.
This value is a little bit higher than the 93.31% that would have been achieved with the previous
clients’ abandonment behavior, as well as adequately higher than the 85% objective. Even growing
in 40% “the clients’ unwillingness”, i.e., decreasing the average waiting time before disconnection
from 2.5 minutes to 1.5 minutes, the impact on the service level was low. Perhaps one should suppose
a higher increase on this indicator, due to the fact that if there are more clients leaving the queue, it
would be more common that the remaining calls could wait less before being handled.
What occurs is that, based to this model, 80% of the disconnected calls return to the queue a few
minutes later, overloading the system once again and not permitting the service level to increase that
much. This type of analysis and conclusion would be mainly impossible through analytical
approaches, which do not consider the abandonment behavior.
If the management is interested only on the service level, it might not become so important to make
great struggles in order to forecast with great accuracy the clients’ average waiting time before
disconnecting the call. This is due to the fact that a not so small change of 40% on this average time
is impotent of impacting completely the service level. If, nevertheless, there is an interest on
monitoring the abandonment rate, too, it is crucial to analyze the impact of the new scenario on this
indicator.
From the 600 calls generated in each replication, 19.3, in average, were disconnected, generating an
abandonment rate of 3.21%, adequately higher than the previous 2.44%.
In order to verify in a more complete way the performance indicators sensitivity related to the average
waiting time before disconnection, the same simulation was replicated a few times by the software,
with different values for this variable. The required outputs were collected to calculate the principal
performance indicators (service level and abandonment rate), which are presented on the following
Table 4.
Table 4. Service level and abandonment rate for different values for the average waiting time
before disconnection, from 00:00 a.m. to 00:30, Sept/15, 12 agents
Source: Table elaborated from the results achieved by the software
As one may notice, the abandonment rate is adequately sensitive to the average waiting time before
disconnection, particularly at lower levels for this variable. As a result, its right consideration looks
to be important on the achievement of accurate results, although the fact that the service level
demonstrates a small sensitivity to changes at the same variable.
The planning manager can also be interested on knowing what would happen if the contracting
company became more demanding in relation to the service level and came to reconsider its concept,
changing it to complement to the percentage amount of clients that waited less than 5 seconds (instead
of 10) before being handled.
During the simulation of this scenario, 579 calls in average were handled, from which 489, in average,
before 5 seconds. The service level for this scenario was, as a result, 84.44%. This value is adequately
lower than the 93.31% achieved with the original definition of service level, and, what is more
important, a little bit lower than the 85% objective settled for extra clients.
206
This comparison shows that a redefinition on the concept of the service level can impact by a not so
small way this performance indicator, something logically expected. It is even possible that, just like
it occurred with the illustrated example, the actual agents’ configuration becomes not sufficient
anymore to give a service level according to the pre-established goal. In this case, it may be relevant
to find out how many additional agents would be necessary to permit this performance indicator to go
back to levels higher than the objective.
In order to figure this out, it is essential to verify if the addition of 1 agent only is good enough to
achieve the goal. In the simulation of this scenario with 13 agents 580 calls were handled, in average,
534 of which (in average) before 5 seconds. The resulting service level was 91.99%, a value higher
than the objective (85%) and as a result higher than the 84.44% achieved with 12 agents, but a little
lower than the original 93.31%.
This demonstrates that the hiring of an additional agent is able of making the service level go back to
the expected objective, but its benefits do not compensate the negative impact on this performance
indicator created by the redefinition of its concept.
The main results obtained by using these analyses (scenarios and sensitivity ones) are summarized
and commented on the following Table 5.
Table 5. Main results achieved by scenarios and sensitivity analyses
Source: Table elaborated by the author
Conclusions
During this research, several simulation models were constructed, completing different real call
centers features and for different possible alternative scenarios, in order to compute performance
indicators and suggest solutions regarding operation sizing. Generally, it was apparent that the
Simulation permits an easy judgment of the impact of changes on the original characteristics of the
operation on the performance indicators. It also permits one performing several sensitivity analyses
related to a few operational parameters.
Particularly, the scenario and sensitivity analyses developed within this paper made us notice how
Simulation can give support to decisions concerning the process of dimensioning a call center, since
the results primarily demonstrated that: (a) it is possible to reduce the agent contingent in some of the
time bands of the day, without much intervention on obtaining the service level objective; (b) not too
huge variations on the received call volumes and on the mean and variability of the handling time can
impact a big deal on the performance indicators, particularly the abandonment rate, indicating the
need to forecast these values with much accuracy; (c) it is possible to improve significantly the
handling performance for preferential clients, without much interference on the quality of the service
for main clients, if an aggregated handling format, with priorities, is adopted; (d) in case the clients
become more impatient and disconnect the calls in a faster way, the impact on the service level would
be small, but very important for the abandonment rate.
207
4.1 Suggestions and Recommendations
In order to structure more correctly the situations to be simulated, it would be interesting if future
researches could make an effort on the direction of finding out (based on the calls historical map) the
correct statistical distribution and the variability for the time among incoming calls and for the
handling time. Several Simulation studies take for granted the key assumption that these times follow
an exponential distribution and develop researches related only to these variables means for each time
band. But the results achieved can be sensitive to the distribution format and to the variability of these
times.
Following this same way of thinking, an empirical research could collect information regarding the
impact caused on the handling time when calls are not answered by the type of agent used to do so,
in a consolidated handling system for different kind of calls. The correct consideration on this impact
tends to generate more accurate results for indicators on call centers with aggregated handling.
Another issue that surely demonstrates a great potential of operational improvement to be analyzed
in future researches is related to the multi-product agent (CAUDURO et al., 2002) [23]. There is a
feeling about being economically favorable to utilize the same agent to handle two or more different
operations at the same time in order to reduce his ineffective time. This could happen in case of
approximately similar operations on which the same agent could work and which present
complementary demand behaviors along the day, the week, or the month. The obtaining of this
supposed economical advantage in terms of cost-benefit could be checked through a well detailed
simulation model, as suggested by Bouzada (2006) [4].
References
1. Grossman, T., Samuelson, D., Oh, S. and Rohleder, T. (2001), En¬cyclopedia of Operations
Research and Management Science, Bos¬ton: Kluwer Academic Publishers.
2. Hawkins, L., Meier, T., Nainis, W. and James, H. (2001), Planning Guidance Document for US
Call Centers, Maryland: Informa¬tion Technology Support Center.
3. Mehrotra, V., Profozich, D. and Bapat, V. (1997), “Simulation: the best way to design your call
center”, Telemarketing & Call Cen¬ter Solutions.
4. Bouzada, M. (2006),
5. Hall, B. and Anton, J. (1998), “Optimizing your call center through simulation”, Call Center
Solutions Magazine.
6. Mehrotra, V. and Fama, J. (2003), “Call Center Simulation Model¬ing: Methods, Challenges and
Opportunities”, Winter Simu¬lation Conference.
7. Anton, J. (2005), “Best-in-Class Call Center Performance: Indus¬try Benchmark Report”, Purdue
University.
8. Brown, L., Gans, N., Mandelbaum, A., Sakov, A., Shen, H., Zel¬tyin, S. and Zhao, L. (2002),
“Statistical analysis of a telephone call center: a queueing-science perspective” (working paper 0312), Wharton Financial Institutions Center.
9. Araujo, M., Araujo, F. and Adissi, P. (2004), “Modelo para seg¬mentação da demanda de um call
center em múltiplas prioridades: estudo da implantação em um call center de telecomunicações”,
Revista Produção On Line.
10. Chassioti, E. and Worthington, D. (2004), “A new model for call centre queue management”,
Journal of the Operational Research Society.
11. Chokshi, R. (1999), “Decision support for call center manage¬ment using simulation”, Winter
Simulation Conference.
12. Klungle, R. and Maluchnik, J. (1997), “The role of simulation in call center management”, MSUG
Conference.
13. Avramidis, A. and L’ecuyer, P. (2005), “Modeling and Simulation of Call Centers”, Winter
14. Klungle, R. (1999), “Simulation of a claims call center: a success and a failure”, Winter Simulation
Conference.
15. Bapat, V. and Pruitte Jr, E. (1998), “Using simulation in call cen¬ters”, Winter Simulation
Conference, p. 1395-1399.
208
16. Mehrotra, V. (1997), “Ringing Up Big Business”, OR/MS Today.
17. Yonamine, J. (2006.
18. Gulati, S. and Malcolm, S. (2001), “Call center scheduling tech¬nology evaluation using
simulation”, Winter Simulation Con¬ference.
19. Paragon (2005), Simulação de Call Center com Arena Contact Center.
20. Miller, K. and Bapat, V. (1999), “Case study: simulation of the call center environment for
comparing competing call rout¬ing technologies for business case ROI projection”, Winter
21. Lam, K. and Lau, R. (2004), “A simulation approach to restructur¬ing call centers”, Business
Process Management Journal.
22. Saltzman, R. and Mehrotra, V. (2001), “A Call Center Uses Simu¬lation to Drive Strategic
Change, Interfaces.
23. Cauduro, F., Gramkow, F., Carvalho, M. and Ruas, R. (2002.
209
IT Outsourcing
Besnik Skenderi1, Diamanta Skenderi2
1
UBT – Higher Education Institution
{besnik_skenderi1, diamanta_skenderi2}@yahoo.com
Abstract Businesses, shareholders and all other interested parties (Custom, Tax Administration and
Customers) require just in time information regarding profit, price, stock and support. Businesses
have responded to those requests with implementation of IT (Information Technology) infrastructure,
but implementation of advanced IT system infrastructure has created cost for shareholder and there
was immediate need to recruit and to train existing staff. With this step, management focus was
oriented in non-strategic processes, and for the implementation and managing of those processes, the
management did not have necessary skills, due to this reason many companies in US, Europe and
Asia have started to outsource those processes. Regarding process of outsourcing Glassman (2010)
concludes, “outsourcing of non-strategic processes eliminates a distraction, enabling management to
focus its distinctive capabilities on areas critical to its strategic success” (p.1).
Keywords: IT, Shareholders, Outsourcing
1. Shareholder Value
IT outsourcing and shareholder values are in correlation, according (Glassman, 2010) “IT outsourcing
has a discernible positive impact on share prices”, (p.1) and according to conducted research in “27
companies undertaking large IT outsourcing initiatives indicates an average gain in shareholder value
of 5.7% and above the general market trend” (Glassman, 2010, p.2).
Shareholder value is increasing for companies that outsource IT, since companies are sharing risk
with vendors and at the same time, companies are benefiting from vendor expertise and knowledge
regarding IT systems. Leijdekker, (2002) states that “One way to outsource receivables management
and feel comfortable about it is to hand off the risk and/or the financing to an outside vendor” (p.27).
Risk is shared in that way since vendors are taking care for all software license fees, data storage and
data processing. Despite this it is in duty of vendor to invest on training for their staff and to provide
proper IT solution systems that are matching requirements of companies.
Businesses benefits from outsourcing their IT because they are using other people and other
companies experience, for one company may be very difficult to implement an IT solution, while the
other one, the implementation of IT solution is a routine job.
Auguste, Hao, Singer, and Wiegand, (2002) are concluding that despite the difficulties, there is a
potential to gain profit by conducting routine operations for other companies.
In another research conducted by Palvia and Xia, and, King, (2011) conclusion is that outsourcing of
IT functions is important component of IT strategy for companies in U.S, Japan and in Western
European countries.
2. Challenges and Examples of It Outsourcıng
Practice and literature confirms that IT outsourcing have positive impact on increasing shareholder
value and Lee, Huynh, Kwok, & Pi, (2003) states that:
“The future prosperity of an organization depends on the quality of its information services. An
organization’s overarching objective in managing its information resources should be to maximize
flexibility and control in order to pursue different options as its circumstances change”, (p.89).
Even that outsourcing will create a profit and it will increase shareholder’s value, this process has its
challenges and risks. Among the risks that can influence successful implementation of outsourcing of
IT are culture differences, labor laws and government rules. One company may outsource it’s IT
210
services in domestic country or in a third party countries, for example many U.S based companies are
outsourcing their IT system processes in India, Philippines and in other less developed countries
where labor force is cheaper but the knowledge and expertise already exists. Before getting involved
in outsourcing process companies should consider cultural differences and time zone since those
factors will affect working culture and schedule. For example, in author’s country (Republic of
Kosovo) is operating 3 CIS, which is IT solution provider company and this company is sub
contracted by many companies in U.S in order to maintain their IT servers.
Many companies had already invested a lot on their hardware, software and in their staff, but the
problem is that they are using those resources only two or three days in a month, only when they need
to produce bills for their customers. Since they have unallocated resources and to gain profit those
companies are offering their services for other companies. Because of government rules and
regulation regarding protection of personal and confidential data, many companies are limited and
cannot outsource their IT services to other vendors. This is a case with Croatian Telecom, this
company is part of German Telecom and German Telecom already has up and running proper billing
system. It seems to be logical to compute all invoices of Croatian Telecom, but since by the law it is
forbidden to send customers data out of the country, Croatian Telecom is forced to invest in hardware
and in software and to compute invoices in Croatia. However, the laws are different in each country,
for example in author’s country (Republic of Kosovo), laws regarding protection of customers are not
imposed by government. Many companies are using this opportunity to outsource some of their
operations, for example KEC (Kosovo Electro energetic Company), is outsourcing billing system in
another country.
But not all the companies are outsourcing their IT systems, some of them had decided to do everything
in house, for example in a company where the author is employed, Post and Telecom of Kosovo
(PTK), all the services are conducted in house and nothing is outsourced. PTK could outsource some
of non-strategic services, like billing system, online customer support and IT help desk. With this
measures PTK shareholder value would be increased and best practices will be implemented.
Sometimes those kinds of decisions could create an opportunity and competitive advantage for
companies. For example, Learning Tree, an international training provider company, in early 1980,
was a hardware producing company based in U.S. This company had started to organize training for
their staff and at the meantime, they had discovered that they are very good in the field of training so
today they are offering only professional training in the field of IT and management.
Conclusıons
Nowadays the success of one company is measured by companies’ ability to compute and to provide
accurate information. Businesses, shareholders and all other interested parties (Custom, Tax
Administration and Customers) require just in time information regarding profit, price, stock and
support. In order to respond to those requests businesses had implemented IT systems. With this step,
management focus was oriented in non-strategic processes, and for implementation and managing
those processes management did not have necessary skills, for this reason many companies in US,
Europe and Asia had started to outsource those processes.
Outsourcing as a concept was introduced in 1990, and after year 2000 many companies had started to
outsource their IT non-strategic processes. With this step, companies are taking a certain amount of
risk, but as the benefit they can focus on their strategy instead losing time and energy on
implementation of different IT systems which are requiring knowledge of latest IT technology and in
a same time they require more human resources.
Shareholder value is increasing for companies that outsource IT, since companies are sharing risk
with vendors and at the same time, companies are benefiting from vendor expertise and knowledge
regarding IT systems
Risk is shared in that way since vendors are taking care for all software license fees, data storage and
data processing. Despite this it is in duty of vendor to invest on training for their staff and to provide
proper IT solution systems that are matching requirements of companies.
Practice and literature confirms that IT outsourcing have positive impact on increasing shareholder
value. Even that outsourcing will create a profit and it will increase shareholder’s value, this process
has its challenges and risks. Among the risks that can influence successful implementation of
outsourcing of IT are culture differences, labor laws and government rules.
211
But not all the companies are outsourcing their IT systems, some of them had decided to do everything
in house and some had discovered that they are very good in different fields.
Generally outsourcing IT creates value and require a willingness to nurture a partnership with another
company, but in the end of day, decision for IT outsourcing is matter of corporate or company culture.
Some companies possess skills and believe that they should do everything in house, while ‘modern’
companies outsource their non-strategic IT operations. IT outsourcing is associated with a certain risk
but in a same time, this process create opportunity for experience exchange and outsourcing can be
used an entry gate to new markets, because during this process companies will gain knowledge
regarding market, labor skills and potential investment opportunities.
References
1
2
3
4
5
Auguste, B. G., Hao, Y., Singer, M., & Wiegand, M. (2002). The other side of outsourcing.
McKinsey quarterly(1), 52-63.
Glassman, D. (2010). IT outsourcing and shareholder value. Stern stewart research, 1-8.
Lee, J. N., Huynh, M. Q., Kwok, R. C.-W., & Pi, S. M. (2003). IT outsourcing evoluation, past,
present and future. Communications of the association for computer machinery, 46(5), 84-89.
Leijdekker, J. (2002). It's time to give credit to outsourcing: how new technology is making it
easier to outsource part of your receivables management. Business credit, 104(2), 25-28.
Palvia, P., Xia, W., & King, R. C. (2011). Critical issues of IT outsourcing vendors in india.
Communications of the association for information systems, 29(11), 203-220.
212
Modelling business and management systems using Fuzzy
cognitive maps: A critical overview
Peter P. Groumpos
Department of Electrical and Computer Engineering, University of Patras, Greece
groumpos@ece.upatras.gr
Abstract. A critical overview of modelling Business and Management (B&M) Systems using Fuzzy
Cognitive Maps is presented. A limited but illustrative number of specific applications of Fuzzy
Cognitive Maps in diverse B&M systems, such as e business, performance assessment, decision
making, human resources management, planning and investment decision making processes is
provided and briefly analyzed. The limited survey is given in a table with statics of using FCMs in
B&M systems during the last 15 years. The limited survey shows that the applications of Fuzzy
Cognitive Maps to today’s Business and Management studies has been steadily increased especially
during the last 5-6 years. Interesting conclusions and future research directions are highlighted.
Keywords: Business Modelling, Management Systems, Fuzzy Cognitive Maps
1. Introduction
Modeling dynamic complex systems can be difficult in a computational sense and today many
quantitative techniques exist.
Well-understood systems may be open to any of the mathematical programming techniques of
operations study. First, developing the model is the most difficult task. It usually requires a great effort
and specialized knowledge from the specific area of interest. This is the case when trying to model
Business and Management (B&M) processes.
Secondly, these systems are often nonlinear, in which case a quantitative model is not easily available
and often may not be possible to be attained. Thirdly these (nonlinear?) systems by necessity involve
feedback controls. In our case, of modelling B&M is not an easy task at all. Fourthly simulating these
systems with real data is almost a next to impossible task to be performed especially today in the
middle of a world economic crisis.
Nevertheless we need to address these problems. Nowadays, due to constant change of business
conditions, flexibility and adaptability are considered significant competitive advantages for an
organization. To achieve this in Business and Management problems, practical problems must be
solved in real time is necessary (Dašić et al. 2011). For the last 40-50 years all problems originated
from Business and Management processes needed experts, in order to be solved in a realistic and costeffective way. Unfortunately, practical problems can arise at any time and experts cannot always be
available. However, their knowledge could be utilized at all times in certain problematic areas with
the use of a well-designed expert system (ES).
Expert Systems constitute the most commonly applied branch of Artificial Intelligence (AI). An ES
is in fact a computational program, which represents and analyzes an expert’s knowledge in a specific
area, so as to advise or solve problems in that field. It emulates human reasoning regarding a problem
by performing a complex process for decision making in a problem domain. Expert Systems, since
their development have been providing us with effective solutions-answers in a variety of problemsquestions that are difficult to handle using other traditional methodologies.
Various ES methodologies (Liao 2005) have led expert systems to overcome their limitations
regarding their rule-based nature. However, there is still a long way to go. ES development continues
to be a time-consuming and expensive task. Additionally, they can solve complex problems in a very
narrow area and it is not always easy for experts to communicate their domain-specific knowledge.
ES are often mistrusted by their users, i.e. managers, who cannot get used to the idea that a computer
program is going to decide instead of them. Managers seem to be fonder of tools for decision
213
facilitation rather than automatic decision making. Thus they have turned to other more sophisticated
methods such as Neural Networks and FCMs.
It is a common truth that the success of business management lies in the consideration and provision
of how a variety of factors interact between them. Today, the number of factors that must be taken
into account for an effective business management has increased significantly, due to the highly
dynamic business environment. Unfortunately, many times managers lack the ability of evaluating all
the related factors, as they use to analyze and assess the impact of two to three factors simultaneously
at best. It is very common for a business system to contain uncertain and fuzzy knowledge, due to the
fact that most knowledge is expressed in terms of cause and effect. In addition, every business
practitioner, i.e. expert, tends to have its individual point of view about effective business
management. Fuzzy Cognitive Maps (FCMs) have come to fill these gaps, as they are best suited for
problems where experts have different opinions about a “correct” answer and they have the ability to
model uncertain and fuzzy knowledge.
FCMs have comprised a tool used for decision facilitation in diverse scientific areas, such as medicine
(Hatwagner 2015) and political science (Tsadiras 2003). The purpose of this study is to review recent
applications of Fuzzy Cognitive Maps in the domain of B&M systems. By doing that, it can be shown
how FCMs can make life for managers a lot easier and it can be derived that FCMs can constitute a
useful tool for decision support in business management, too.
This paper has been organized in the following way. Section 2 gives a brief overview of Fuzzy
Cognitive Maps, while Section 3some limitations of FCMs. Section 4 presents recent applications of
FCMs to various areas of business and management. In Section 5 the survey of FCMs been used in
B&M is provided on a table and a short discussion on the survey results is given. Finally, in Section
6, conclusions and future research topics are outlined.
2. A Brıef Overvıew of Fuzzy Cognıtıve Maps
Fuzzy Cognitive Maps (FCMs) is a modeling technique, arising from the combination of Fuzzy Logic
and Neural Networks. Compared to conventional expert systems, FCMs have considerable
advantages; they are dynamic feedback systems (Taber 1991) and they can represent structured
knowledge very easily, providing the higher level of abstraction needed in many cases (Lee & Ahn
2009). FCMs constitute a useful tool, with which we take advantage and quantify the accumulated
knowledge obtained through years of observing the operation and behavior of complex systems
(Papageorgiou et al. 2003). Thus we can claim that Fuzzy Cognitive Maps are fuzzy structures that
strongly resemble neural networks, and they have powerful and far-reaching consequences as a
mathematical tool for modeling dynamical complex systems.
The term of “fuzzy cognitive map” was first used by Kosko (Kosko 1986) to illustrate a graphically
signed directed graph with feedback, consisting of nodes and weighed arcs. The FCM’s nodes
represent the concepts used for describing system behavior. These nodes are interconnected by signed
and weighted arcs, standing for the existing causal relationships between the concepts. In other words,
the arcs describe the variation on the value of one concept when the value of an interconnected concept
is being altered.
When concept Cj influences concept Ci, there is a wji arc which can take any value between -1 and 1,
quantifying this way the correlation between the two concepts. wji can be:
Positive (wji>0) when there is a positive causality between C j and Ci, i.e. an increase/decrease in the
value of Cj brings an increase/decrease in the value of C i respectively.
Negative (wji<0) when the causality is negative and increase/decrease of the value of Cj causes a
decrease/increase of the value of Ci.
Zero (wji=0) when there is no influence of concept Cj to concept Ci.
The bigger the absolute value of wji it is, the stronger the influence of C j to Ci will be, in a direct
(positive causality) or an inverse way (negative causality).
In a conventional FCM, the value of each concept is computed, taking into account the influence of
other concepts to the specific concept (Groumpos 2010), by applying the following mathematical
procedure:
214
N
Ai ( k 1)  f (k2  Ai( k )  k1   Aj( k )  w ji ).
(1)
j i
j 1
In (1), Ai(k+1) is the value of concept Ci at time k+1, Aj(k) is the value of concept Cj at time k, wji is the
weight of interconnection between concepts Cj and Ci and f is the sigmoid threshold function.
Fig. 1. A Simple Fuzzy Cognitive Map
A more detail mathematical presentation of FCMs is provided on Glykas (2010) and Groumpos
(2010)
3. Some Limitations of FCM
In reviewing the literature, for the last 30 years it has been realized that FCMs, have strong and weak
points. They are flexible, adaptable and they can model very well semi-structured or unstructured
knowledge. Nevertheless, they contribute to the problem solving process only by evaluating the
alternatives of a scenario, thus, not absolving the manager from making the actual decision. The
preceding features suggest that FCMs cannot be utilized in all kinds of problems. They are preferred
in unstructured cases, where they can provide managers with very good results without the ethical
implication of human replacement. Another short come of FCMs is that present models and
algorithms cannot guarantee convergence to a steady value. In addition without learning algorithms
all different initial conditions give the same final value of the FCM if and when it converges. Another
limitation is the way that experts are utilized in formulating the FCM model and how different expert’s
knowledge is taking into consideration. For B&M processes needed reliable and real data are not
easily available.
Thus although FCM provide a powerful tool for both traditional experts and non-traditional experts
to model complex systems, evaluate structural differences between the knowledge held by groups and
individuals, and functionally determine the dynamic outcome of this understanding, there are
still issues regarding the interpretation of FCMs as artefacts of individual knowledge and group
beliefs.
In this paper, a theoretical background in order to inform the collection and make interpretation of
FCM as repre- sentations of shared knowledge when individual FCMs are aggregated together,
compared across individuals within the context of group interaction, or created collectively by
individuals within a group context and especially in B&M.
4. Applıcatıons of FCMs in Business and Managment
Considering the abovementioned favorable features of the FCMs, and the overall increase of FCMrelated studies (Glykas 2010, Groumpos 2010) there has been also an increasing amount of literature
on business and systems management in the last 30 years. In this work, an attempt is made to carefully
review FCM research studies related to business and management and illustrate some interesting
applications of them.
215
Several studies have succeeded in measuring and assessing business performance using FCM. As
early as 1989 and then 1992 Zhang et all. used FCM on decision analysis for business processes. Also
in 1997 D. Kardaras, and G. Mentzas used FCM to analyze Business Performance Assessmen. In
another work, Glykas (2013) FCMs were implemented in strategy maps, eliminating their drawbacks
and providing them with competitive advantages in terms of flexibility and simulated scenarios.
Results of case studies showed that the developed system could emulate effectively experts’
predictions. Chytas et al. (2011) managed to develop a new proactive balanced scorecard
methodology with the aid of FCMs, which were used for quantifying the imprecise causal
interrelationships of key performance indicators, so as to adjust performance goals. FCMs are used as
a performance modeling tool for the implementation of business performance re-engineering
methodology, in order to simulate the operational efficiency of complex, imprecise functional
relationship and quantify the impact of BPR activities to the business model (Xirogiannis & Glykas
2004).
FCMs have found a great applicability in the planning process, too. Lee et al. (2013) employed FCMs
to industrial marketing planning. By integrating agent and FCM technology, they managed to
overcome the conventional FCMs’ limitations in marketing planning. Hence, experts’ opinions from
different functional departments were integrated and complex, ambiguous causalities among the
related variables were quantified, allowing this way a systematic what-if analysis to be carried out, in
order to compare various scenarios. Comparison and evaluation of different scenarios is done in
another work too (Lopez & Salmeron 2014), in which FCMs are applied to enterprise resource
planning, modeling maintenance risks. With the help of FCMs, ERP practitioners highlight the most
important factors and are able to handle the maintenance risks more effectively. Kardaras and
Karakostas (1999) used FCMs as a supplement to the strategic planning of information systems. This
way, it could be much easier for planners to label specific IT projects and evaluate their impact on an
organization. Regarding business informatics, an interesting work was that, which proposed
augmented FCMs for modeling LMS critical success factors, in order to discern the necessary
activities for success (Salmeron 2009)
FCMs have also addressed the growing need to assess investment decision-making processes. Irani et
al. (2014) managed to shed light upon the often cloudy evaluation of IS investments by identifying,
classifying and correlating the factors that affected more the IS investment evaluation with the related
knowledge components. As a result, an analysis of knowledge-based decisions in the IS evaluation
investment area took place. In 2004, Koulouriotis assigned FCMs to emulate investors’ reasoning
process, as a means for stock prediction.
Human Resources Management has also been implemented with the fuzzy cognitive mapping
technique. Xirogiannis et al. (2008) attempted to model the operational efficiency of HRM activities
and their impact to shareholder satisfaction. Thus, the effects of HR practices to the overall
shareholder value were clearly illustrated. In another work, Oleyaei-Motlagh and Bonyadi-Naeini
(2014) investigated HRM influence to Six Sigma projects implementation. The critical factors were
identified, so managers knew where should focus, in order to achieve better organizational
performance as far as HRM is concerned. Last but not least, FCM applications, such as modeling core
value systems (Macedo & Camarinha-Matos 2013) and relationship management (Kang et al. 2004)
enable possible detection of conflicts among colleagues, doing in this manner, the work of HRM much
easier. As far as e-business is concerned, Miao et al. (2007) integrated Intelligent Agents into Fuzzy
Cognitive Mapping, creating the tool of Fuzzy Cognitive Agents, in order to support e-commerce
applications. That new approach enabled personalized recommendations based on the online user’s
preferences, common preferences of other users in general and expert’s knowledge. Hence, they
addressed the difficulty of many users, who cannot determine what they really want, when visiting
electronic commerce web-sites. Also, Lee and Ahn (2009) combined FCMs and structural equation
modeling for control design support of an e-commerce web based system (ECWS), in order to achieve
high ECWS performance. The FCM’s fuzzy edges facilitated the representation of environmental
factors, controls, implementation, and performance in ECWS, while the structural equation modeling
defined the complex causal relationships among factors. In another publication, Xirogiannis and
Glykas (2007) used FCMs as a means for a causal representation of dynamic e-business maturity
indicators. They showed how FCMs can work as a supplement to strategic change projects, by
modeling complex strategic models and quantifying the impact of strategic changes to the overall ebusiness efficiency.
216
Other worth to be mentioned applications follow. Nasserzadeh et al. (2008) used FCM modeling, in
order to assess customer satisfaction, a competitive advantage in today’s societies, in the sector of
banking industry. The Delphi methodology, which included feedback reports for everyone, was
exploited for easier consensus among the experts. The resulting FCM was capable of evaluating a
bank’s capacities for attracting new customers and increase customer satisfaction rate. Irani et al.
(2009) used fuzzy cognitive mapping to investigate the relationship between knowledge management
and organizational learning. They showed that a relationship between them does exist and the
constructed FCM helps them to identify factors, with which a company could evolve to a learning
organization. Wang et al. (2011) used structural equation model as a supplement to FCM, in order to
define the complex causal relationships among factors in a hotel environment. The constructed FCM
enlightened the often cloudy context of building competitive advantages in the hotel industry and it
enabled a systematic what-if analysis to be carried out, in order to compare various possible scenarios
in reality. FCMs were also applied with success in the complex sector of retail industry. They led to
a better implementation of collaborative planning, forecasting and replenishment (CPFR) approach,
as they highlighted the related factors for CPFR success (Büyüközkan & Vardaloğlu 2012).
5. Discussion on Survey Results
In section 4 a small and limited number of FCMs applications to B&M were presented. Indeed it was
difficult to find related studies prior to 1990s and even prior to 2000s where FCMs were used in B&M
processes. This is reasonable and understandable since the notion of FCMs was introduced by Kosko
in 1986. Extensive studies of FCMs took place late in the 1990s and early in 2000s. However
surprisingly FCMs were used extensively in B&M studies from the early 2000s as will be shown
next.
Through a partial but not as exhaustive as needed extensive survey and classification of FCM
published studies across the internet, last 15 years FCM applications in B&M are provided in Table
1. As been expected the survey could not and did not covered all possible application of FCMs in the
broad scientific area of B&M. An effort was made to carefully review the first 6-monts of 2015, a
task not so easy given that many studies, conferences and book chapters have not yet been recorded
in any data base system. This should be a future research study. However Table 1 gives us some
interesting statistical results. There has been a steady increase in the number of FCM-related studies
in this field, especially the last 5-6 years. In particular, FCM studies regarding Business &
Management in 2014 have been increased by almost a magnitude of ten (10) compared with those
fifteen (15) years ago. Similarly for the papers and/or book chapters Table 1 shows a considerable
increase in numbers in the last 5 years compared to the period of 2000-2010. Considering the above
mentioned facts , it is clear that FCM application in business and management has gained a
considerable interest over the last five years.
It is of interest to point out that during the first 7 months of 2015 there many studies and papers that
have been in the literature in which FCMs are used extensively in studies on B&M systems. For
example in Uden et al. (2015) contains the refereed proceedings of the 10th International Conference
on Knowledge Management in Organizations, KMO 2015, held in Maribor, Slovenia, in August 2015.
The theme of the conference was "Knowledge Management and Internet of Things" in which more
than 200 papers were presented and in most of them the central point was KM in organizations and
the problems were treated using extensively Fuzzy Logic, Intelligent Systems and FCMs. The
interested researcher should search through the “FCM in B&M” and would be surprised with the
results. FCMs have been extensively in B&M studies the last 2-3 years.
217
Table 1. Recent FCM Studies in Business &Management
Number of FCM-related studies and papers in
Business & Management
Year
Studies
Journals &
Book
Chapters
Conferences
2000
2
0
2
2001
0
2
3
2002
3
6
2
2003
5
5
4
2004
12
2
8
2005
3
0
6
2006
9
5
4
2007
9
3
6
2008
12
5
10
2009
20
16
7
2010
20
13
10
2011
15
12
8
2012
17
2013
22
18
12
2014
28
22
15
2015
(7 months)
5
7
15
14
9
An earlier paper by Darlas and Groumpos (2014) covered only the survey aspects of using FCMs in
B&M systems and for a smaller period of time, the10 years 2003-2013. This study covers more than
15 years; it is a little more extensive thus the provided table of Darlas and Groumpos (2014); here
also differs on the reported results even for the period 2003-2013. As a matter of fact this study has
covered the whole period since 1985 almost 30 years. Indeed the earliest study that was found in
which FCMs were used on B&M studies is that Zhang et all. (1989). The important finding here is
that just 3 years after Kosko (1986) has introduced for first time the concept of Fuzzy Cognitive Maps
(FCMs), this new method has been used on Business studies. Thirty years later in 2010, more than 40
related “studies” (studies, papers, and book chapters) have been identified on this limited study to
relate FCMs and B&M studies. In the first seven (7) months of 2015 on the limited sources that have
been searched more than 20 studies are covering this broad aspect of applying FCMs in Business and
Management studies. This trend will continue with a steady increase in using FCMs in B&M
processes.
Conclusıons and Future Research
Business and Management problems are characterized by controversial theories and mathematical
solutions, with complex equations and formulas. These difficulties have led to an extensive use, for
the period up to 2000, of Expert Systems (ES) and Intelligent Systems (IS) to address problems of
this domain, thanks to their capability of taking advantage of experts’ domain-specific knowledge and
emulate their inference in a very narrow area. However they were not able to provide satisfactory
solutions especially after 2000 when the international B&M systems became so complicated due to
their globalization. This study was set out to demonstrate the various applications of FCMs to B&M.
FCMs have been proven through the limited literature, as a very useful method and tool to identify
218
and assess the key-factors of a variety of B&M issues. A table shown the use of FCMs in B&M
systems has been presented. The results of this review, does not show the whole magnitude of the use
of FCMs in B&M systems. However it demonstrates a positive trend in using FCMs in B&M for the
last 10 years.
However, FCM technique cannot act as a panacea for solving B&M problems as it depends upon
several assumptions, such as the existence of only symmetric or monotonic causal relationships. In
addition the globalized economy presents totally new, challenging and controversial problems
especially when there are many and different economic conflict of interest. A lot of effort has been
made for conventional FCMs to overcome their drawbacks either by using various supplements or by
advancing FCM theory and learning. Future research directions could include: develop new B&M
systems for special purposes using FCMs use real data to validate the new models. Study the behavior
of B&M systems using Existing FCM methods and tools. We need to research more, how studying
the historical knowledge of B&M processes leads us to wise decisions in the future. See plenary paper
(Groumpos 2015) of this conference. Despite their drawbacks though, the increasing number of
FCMs’ applications not only in various aspects of business, but also in diverse scientific fields, seems
quite promising.
References
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Büyüközkan, G. and Vardaloğlu, Z. (2012). Analyzing of CPFR success factors using fuzzy
cognitive maps in retail industry. Expert Systems with Applications, 39(12), pp.10438-10455.
Chytas, P., Glykas, M. and Valiris, G. (2011). A proactive balanced scorecard. International
Journal of Information Management, 31(5), pp.460-468.
Darlas, O.D and Groumpos, P. P. (2014)A Survey on Applications of Fuzzy Cognitive Maps in
Business and Management. In Proc. of the International Conference
“Information Technologies for Intelligent Decision Making Support” May 2014. Ufa, Russia
Dašić, M., Trajković, S. and Tešanović, B. (2011). The necessity of using expert systems in
strategic decision making. International Journal of Economics & Law, 1(1), pp.27-35.
Glykas, M. (2010). Fuzzy cognitive maps. Berlin: Basic Theories and their Application to
Complex Systems. In: M. Glykas, ed., Fuzzy Cognitive Maps: Advances in Theory,
Methodologies, Tools and Applications, Springer-Verlag Berlin Heidelberg.
Glykas, M. (2013). Fuzzy cognitive strategic maps in business process performance
measurement. Expert Systems with Applications, 40(1), pp.1-14.
Groumpos, P.P. (2010). Fuzzy Cognitive Maps: Basic Theories and their Application to
Complex Systems. In: M. Glykas, ed., Fuzzy Cognitive Maps: Advances in Theory,
Methodologies, Tools and Applications, Springer-Verlag Berlin Heidelberg, pp.1-22.
Groumpos, P. P. (2015),The Need for Wise Decision Making Support Systems (WDMSS) In
Developing Future Intelligent Systems. (this conference, TECIS 2015)
Hatwagner, M. (2015). Introduction of Modeling Complex Management Systems using Fuzzy
Cognitive Map. In: 7th International Conference on Information Technology ICIT. pp.508-514.
Irani, Z., Sharif, A., Kamal, M. and Love, P. (2014). Visualising a knowledge mapping of
information systems investment evaluation. Expert Systems with Applications, 41(1), pp.105125.
Irani, Z., Sharif, A. and Love, P. (2009). Mapping knowledge management and organizational
learning in support of organizational memory. International Journal of Production Economics,
122(1), pp.200-215.
Kang, I., Lee, S. and Choi, J. (2004). Using fuzzy cognitive map for the relationship
management in airline service. Expert Systems with Applications, 26(4), pp.545-555.
Kardaras, D. and Karakostas, B. (1999). The use of fuzzy cognitive maps to simulate the
information systems strategic planning process. Information and Software Technology, 41(4),
pp.197-210.
Kardaras, D., and Mentzas, G. (1997). Using fuzzy cognitive maps to model and analyze
business performance assessment. In Prof. of Int. Conf. on Advances in Industrial
Engineering–Applications and Practice II , pp. 63-68.
Kosko, B. (1986). Fuzzy cognitive maps. International Journal of Man-Machine Studies,
24(1), pp.65-75.
219
17 Koulouriotis, D. (2004). Investment analysis & decision making in markets using adaptive
fuzzy causal relationships. Operational Research, 4(2), pp.213-233.
18 Lee, S. and Ahn, H. (2009). Fuzzy cognitive map based on structural equation modeling for the
design of controls in business-to-consumer e-commerce web-based systems. Expert Systems
with Applications, 36(7), pp.10447-10460.
19 Lee, K., Lee, H., Lee, N. and Lim, J. (2013). An agent-based fuzzy cognitive map approach to
the strategic marketing planning for industrial firms. Industrial Marketing Management, 42(4),
pp.552-563.
20 Liao, S. H. (2005). Expert system methodologies and applications—a decade review from 1995
to 2004. Expert Systems with Applications, 28(1), pp.93-103.
21 Lopez, C. and Salmeron, J. (2014). Dynamic risks modelling in ERP maintenance projects with
FCM. Information Sciences, 256, pp.25-45.
22 Miao, C., Yang, Q., Fang, H. and Goh, A. (2007). A cognitive approach for agent-based
personalized recommendation. Knowledge-Based Systems, 20(4), pp.397-405.
23 Oleyaei-Motlagh, S. and Bonyadi-Naeini, A. (2014). Identifying the role of human resource
management in increasing performance and implementation of six sigma projects using fuzzy
cognitive maps. 10.5267/j.uscm, 2(3), pp.179-190.
24 Papageorgiou, E. (2011). Review study on fuzzy cognitive maps and their applications during
the last decade. In: 2011 IEEE International Conference on Fuzzy Systems. IEEE Computer
Society, pp.828-835.
25 Papageorgiou, E., Stylios, C. and Groumpos, P. (2003). Fuzzy Cognitive Map Learning Based
on Nonlinear Hebbian Rule. In: AI 2003: Advances in Artificial Intelligence Lecture Notes in
Computer Science, Springer Berlin Heidelberg, pp.256-268.
26 Reza Nasserzadeh, S., Hamed Jafarzadeh, M., Mansouri, T. and Sohrabi, B. (2008). Customer
Satisfaction Fuzzy Cognitive Map in Banking Industry. In: Communications of the IBIMA, 2
(21). pp.151-162.
27 Salmeron, J. (2009). Augmented fuzzy cognitive maps for modelling LMS critical success
factors.Knowledge-Based Systems, 22(4), pp.275-278.
28 Taber, R. (1991). Knowledge processing with Fuzzy Cognitive Maps. Expert Systems with
Applications, 2(1), pp.83-87.
29 Tsadiras, A., Kouskouvelis, I. and Margaritis, K. (2003). Using Fuzzy Cognitive Maps as a
Decision Support System for Political Decisions. In: Advances in Informatics, Springer Berlin
Heidelberg, pp.172-182.
30 Uden, L., Heričko, M. and Ting, I. ed., (2015). Knowledge Management in Organizations: 10th
International Conference, KMO 2015. Maribor, Slovenia, Proceedings (Vol. 224). Springer.
31 Wang, C., Chen, S. and Chen, K. (2011). Using fuzzy cognitive map and structural equation
model for market-oriented hotel and performance. African Journal of Business Management,
5(28), pp.11358-11374.
32 Xirogiannis, G., Chytas, P., Glykas, M. and Valiris, G. (2008). Intelligent impact assessment of
HRM to the shareholder value. Expert Systems with Applications, 35(4), pp.2017-2031.
33 Xirogiannis, G. and Glykas, M. (2007). Intelligent modeling of e-business maturity. Expert
Systems with Applications, 32(2), pp.687-702.
34 Xirogiannis, G. and Glykas, M. (2004). Fuzzy Causal Maps in Business Modeling and
Performance-Driven Process Re-engineering. Methods and Applications of Artificial
Intelligence Lecture Notes in Computer Science, pp.334-351.
35 Zhang W.R, Chen S.S, Bezdek J.C,(1989) Pool2: A generic
36 system for cognitive map development and decision
37 analysis, IEEE Transactions on Systems, Man, and
38 Cybernetics 19 (1)
39 Zhang, W., Chen, S., Wang, W. and King, R. (1992). A cognitive-map-based approach to the
coordination of distributed cooperative agents. IEEE Transactions on Systems, Man, and
Cybernetics, 22(1), pp.103-114.
220
E-customer relationship management in insurance industry
in Albania
Evelina Bazini
University of Vlora “Ismail Qemali“, Faculty of Economy
Albania
evelina.bazini@univlora.edu.al
Abstract. E- Customer relationship management is an issue that every company, large or small must
take in some way. Handled well, a CRM strategy can deliver significant benefits for companies and
customers. Interaction with customers, in particular, has been enhanced and organizations who wish
to remain competitive have started to implement CRM programmes and techniques in order to
develop closer relations with their customers and to develop a better understanding of their needs. At
the same time, the use of e-commerce techniques in CRM allows insurance organizations to identify
customers, monitor their habits and use of information, and deliver them improved information and
services according to their recognized needs and buying behavior. This paper looks at new
developments in e-CRM in the insurance industry.
Keywords: e-customer relationship management, insurance industry, ICT.
1 Introduction
Over the last two decades, the insurance industry has experienced remarkable changes in its
organizational structure due to the increased use of the Internet in business processes. The advent of
the Internet, along with increasing globalization and growing competition have brought fundamental
changes in the way information about insurance products is distributed and delivered. The Internet
and other ICTs offer extraordinary opportunities for innovation in customer relations management.
These technologies may empower customer choice and offer unprecedented amounts of information
about customer needs and desires. They offer new means for customers to find product and service
information, to negotiate purchases, to make enquiries about the conditions of sale and support for
their purchases, and to track deliveries and requests. For businesses, e-CRM provides new ways to
recruit customers and retain their loyalty, to customize services and provide personalized products
and services, to identify emerging patterns of demand in style and to enhance links with suppliers to
meet these demands. E-CRM is a good candidate for a systematic examination of organizational and
technological innovation because of the diversity and potential impact of the opportunities it provides.
2. Literature review
2.1. Insurance services2
According to the Encyclopaedia Britannica, insurance is “a contract for reducing losses from accident
incurred by an individual party through a distribution of the risk of such losses among a number of
parties.” The definition goes on to say: “In return for a specified consideration, the insurer undertakes
to pay the insured or his beneficiary some specified amount in the event that the insured suffers loss
through the occurrence of a contingent event covered by the insurance contract or policy. By pooling
both the financial contributions and the ‘insurable risks’ of a large number of policyholders, the
insurer is typically able to absorb losses incurred over any given period much more easily than would
the uninsured individual” (Encyclopaedia Britannica, Micropaedia, 1987, p. 335). A briefer definition
of insurance as a phenomenon is “the practice of sharing among many persons, risks to life or property
221
that would otherwise be suffered by only a few. This is effected by each person paying a sum of
money called a premium which, with those paid by all the others, is put into a ‘pool’ or insurance
fund, out of which money is paid to the few who suffer loss” (Longman Dictionary of Business
English, 1989). The policyholder thus pays someone else a premium to bear his or her risk, knowing
that a possible future loss will be compensated for according to the premium paid. If lucky, the
policyholder will never have to experience the tangible results of the service of reduced risk during
the contracted policy period. On the other hand, the policyholder maintains a certain uncertainty
towards the service that he or she pays for, something that adds to the peculiarity of insurance services.
2.2. Relationship marketing in an insurance environment2
A relationship marketing approach allows the insurance marketer to offer a product in response to
needs triggered by the customer and based on experience and information gathered over time. Sales
and profitability can be dramatically increased because the more a marketer knows about a customer
the more effectively the customer can be approached with appropriately targeted products (Harrison,
1993).One of the major themes in relationship marketing, as well as a key to profitability, is to develop
long-term relationships with customers. This involves the ability to retain customers, and is in turn,
dependent on agents possessing the “right” characteristics. What is “right” varies depending on
whether the customer is an individual or a company. In insurance language these two types of
customers may be called personal lines policyholders and commercial lines policyholders respectively
2.3 E-customer relationship management7
E- CRM describes the broad range of technologies used to support a company’s CRM strategy. It can
be seen to arise from the consolidation of traditional CRM with the e-business applications
marketplace. Lee Kelley et al. (2003, p. 24) highlight the relative lack of literature in this domain and
suggest as a working definition that e-CRM refers to ‘the marketing activities, tools and techniques
delivered via the Internet which includes email, world wide web, chat rooms, e-forums, etc., with a
specific aim to locate, build and improve long term customer relationships to enhance their individual
potential’. Typically electronic and interactive media such as the Internet and email are seen as playing
the most significant role in operationalizing CRM as they support effective customized information
between the organization and customers. However, e-CRM can also include other e-technologies and
new e-channels including mobile telephony, customer call and contact centers and voice response
systems. The use of these technologies and channels means that companies are managing customer
interactions with either no human contact at all, or involving reduced levels of human intermediation
on the supplier side (Anon, 2002). The emergence of mobile commerce has led to the introduction of
new products, new ways of selling products to customers and new learning curves for companies in
terms of how to manage interactions with customers (Wright et al., 2002).
3. The Albanian insurance market8
Insurance affects significantly economic growth because through risk management instruments it
promotes efficient investments. The Albanian Insurance Market is a new market, dynamic and with
continuous emerging developments. Insurances in Albania are relatively a young field though their
inception dates in 1900. The well-known Londoner insurance company Lloyd’s had branches in
Durresi, Vlora and Shengjin. Insurance market in Albania has passed through several stages:
First stage – dates on the Second World War, when all insurances activities where carried out by
foreign companies (Londoner, Italian & French). Italian companies acting in Albania were
Assicurazione General, Reunione Adriatica, Fonderia Di Firencia Tirana, Instituto Italiana di
Previsone, and Roberto Alegro Vlore etc. Italian companies covered insurance activities related with
natural disasters in premises, industries and life: Introducing for the first time the insurance concept
in the albanian life.
Second Stage – lasts from 1944 to 1991. In that time during a meeting the Ministers Council
determined the obligatory insurance of government property, while for the private property, insurance
was enforced by law. Objectives, risks, premiums, regulation and other conditions related with
insurances were approved by the Finance Ministery over the proposals of Savings & Insurance
222
Institute. During this stage it is impossible to discuss about real insurance market in Albania. The
economy was centralized, so every gathered premium was used according to a plan by the state. It
was impossible for the institution to have its own investment portfolio.
Third stage – from July 1991 up to no. The big changes that happened in the Albanian life had their
reflections in the economic life, and following this also in the insurance field.
In 31 July 1991, the ministers’ council approved the law nr 7506, according to which, the Savings
and Insurance institute, would split up in two different institutions: Savings Bank and Insurance
Institute named also INSIG. INSIG the first Albanian insurance company was created as a commercial
enterprise and acted in compliance with market economy rules. A very important step in reformation
and liberalization of the insurance system in Albanian played also the law nr 8081 dated on 07.03.1996
“For the insurances and/or reinsurances activities’.
4. Methodology of research
The data presented in this paper come from an exploratory study involving detailed interviews with a
variety of leading industry actors in the insurance industry using a structured interview questionnaire.
The questionnaire was designed to serve as the basis for qualitative rather than quantitative research.
Specifically, it was intended to collect qualitative information about the status of e-CRM and trends
in the tourism service industry and also to embrace the views of those involved directly in CRM. The
questionnaire results were processed anonymously and the interviewee data have been treated
confidentially. The research methodology enabled the collection of both qualitative and quantitative
information about a number of trends surrounding the emergence of e-CRM in the insurance industry.
Overall, competition within the insurance industry is becoming more and more intense.
Research findings
Although most interviewees stressed that it is too early to assess the impact of e-CRM methods on
customer relations, they felt that their companies' performance in relation to customer retention and
satisfaction had improved as a result of their online presence. Several barriers and constraints were
identified in the adoption of e-CRM including the maturity of the market to which companies address
their products and services, the financial resources required, and the prevailing structures and modes
of practice within organizations which inhibit the establishment of new business routines.
These findings suggest that although the majority of surveyed companies value e-CRM and have
started to commit resources to relevant initiatives, few companies have actually been engaged in
mature, long term planning regarding e-CRM proper adoption and use. However, the real challenge
for insurance companies in order to realize productivity gains is the automation of their back office
operations. The challenge therefore is not to implement technology-based CRM solutions but to
establish the associated organizational and cultural (customer-centered service provision) change
within companies that would allow CRM to contribute to productivity and profitability.
Functional changes associated with e-CRM adoption - The majority of companies in the sample were
in the process of unravelling their e-CRM development plans and were not prepared to comment on
the changes in the functions of their organization as they had not been in a position to fully exploit
the expected opportunities. E-business methods enable a company to grow without developing its
physical distribution network. At the same time, all companies papered that customer relations have
become stronger and more 'personal' with the use of e-CRM, as the technology allows them to contact
customers individually in a fast and cheap way and to receive frequent feedback from them.
Gaining and retaining customers - The interviewees' responses to alternative e-business methods for
identifying, gaining and interacting with customers. These may include online promotion campaigns,
maintaining a customer database, customer profiling, setting up interactive questionnaires, and
tracking the choices of customers while they surf on the website. The responses show that companies
are particularly keen on maintaining customer databases and less keen on setting up interactive
questionnaires, which as suggested are often treated as a nuisance by the customers.
The ability to collect, combine, maintain, and process data from a multiplicity of sources is central to
gaining a richer view of customers, their needs, and the value they can generate for the company.
223
Bringing together data from different sources is a critical task for insurance companies that wish to
develop a better understanding of their customers. A model that has been proposed for the evolution
of customer databases involves three stages: In the first stage, data are stored in electronic format by
each department of the organization and/or by type of application without the possibility for crosschecking and information sharing. The next stage involves the development of 'links' between the
different databases so that when a customer's record is updated in one system, it automatically passes
the information onto the others. The third and more advanced stage involves the development of a
data warehouse system which has sufficient intelligence to perform more complicated tasks such as
analyzing customer data and assessing the value of a customer.
Support customer information needs - Interviewees agreed that Internet websites provide richer and
more-up-to-date information than any other physical distribution channel. For the majority of
companies the web serves as a primary source of information and marketing tool for their services.
E-catalogues are cost-effective, can be more detailed than paper catalogues, and are friendly to use
even for customers with basic Internet skills. Some interviewees expressed the view that soon they
will replace paper-printed brochures.
Organizational impact of e-CRM adoption - The adoption of e-business methods in customer relations
signaled organizational changes in most of the companies in the sample. Some companies established
a special department within the organization. The department usually consists of a working team of
people dedicated to set up and maintain the website and the sales over the Internet.
Although most interviewees stressed that it is too early to assess the impact of e-CRM methods on
customer relations, they felt that their companies' performance in relation to customer retention and
satisfaction had improved as a result of their online presence.
Perceived barriers and future trends - In their early steps toward e-CRM, the insurance companies
that took part in the research identified several barriers and constraints. These mainly related to:
- The maturity of the market to which they address their products and services
- The financial resources required, and
- The prevailing structures and modes of practice within organizations.
An additional inhibiting factor identified by company representatives was the size of investments
required for the introduction of e-CRM given that the industry generally operates with low margins.
A customer-focused strategy driven by the use of new technology requires large investments that need
to be justified to shareholders and compensated in a reasonable period of time. Apart from the
conflicts that arise with traditional distribution networks of insurance agents, a further constraint
relates to the prevailing organizational structure within companies. Companies were not always ready
to adopt and exploit new ways of doing things. Some interviewees papered resistance to change by
some departments, which slowed the adoption process. A minimum level of agreement on a common
vision is necessary to get things started and an orchestrated effort is required so that internal conflicts
between new and traditional sale methods are eliminated.
The barriers identified above have made many companies take a step-by-step approach in
implementing e-business methods. Such a cautious strategy is expected to take into account a
company's identity and image with customers as well as an assessment of possible impacts and results.
Despite some skepticism, the majority of companies responded positively as to whether they intend
to extend e-business in the organization within the next five years. An argument put forward is that
'what is acceptable by e-commerce methods now, will expand as clients get used to, and confident
with the medium so this may present scope for expansion'. Others considered further expansion of ebusiness initiatives to be a strategic choice for their companies as to the way to become leading players
in the online market. Only a small number of interviewees argued that such plans fall within normal
business practice.
Conclusions and recommendations
The research findings indicate that companies in the insurance industry in Albania are just beginning
to experiment with e-CRM applications and to explore the business opportunities they entail. So far,
these have remained largely untapped as witnessed by the fact that traditional sales methods still
account for over 90% of their revenues. The competitive pressure to keep up with the new business
environment appears to be a major factor driving e- CRM developments in the industry.
224
Traditional and new players in the insurance industry see the Internet more as an efficient tool of
communication and interface with customers than as a fundamentally new way of doing business.
The availability of services online offers opportunities to companies in the insurance business to
achieve cost reductions in operating costs and opens up an effective and direct communication channel
with existing and potential customers. Managing the relationships with the network of independent
agents, the purchasing groups, and their own off-line and online distribution networks emerged from
the interviews as a fundamental problem suppliers in the insurance industry are facing. An alternative
solution has been proposed that would reconcile the traditional with the new distribution channels and
would turn the Internet into an ally instead of an enemy of insurance agents. For this to happen, the
latter need to understand and value the complementary nature of online selling to traditional business
practices.
The real challenge therefore is not to implement technology-based CRM solutions but to establish the
necessary technological (data-warehouses and customer intelligence systems), organizational (data
integration, inter-departmental communication, links with distributors and resellers) and cultural
(customer-centered service provision) change within companies that would allow CRM to contribute
to productivity and profitability.
References
1. Ahmed Mohiddin. (2002). The challenges of e-democracy. Opportunities and risks.
2. Bazini, E. Elmazi, L. & Sinanaj, Sh. (2012). Importance of relationship management in the
insurance business in AlbaniaProcedia - Social and Behavioral Sciences. Volume 44, Pages 1-482
(2012). ISSN: 1877-0428.
3. Bloch, M. and Y. Pigneur and T. Steiner (1996). The IT-enabled Extended Enterprise:
Applications in the Tourism Industry. Stefan Klein et al. Information and Communication
Technologies in Tourism. Ohrid : Macedonia
4. Buhalis, D. & M.C.Licata (2001).The Future of e-Tourism Intermediaries. Journal of
Tourism Management, 23(3): 207-220.
5. Buhalis, D. (1998). Strategic Use of Information Technologies in the Tourism Industry. Jornal of
Tourism Management, 19(5): 409-421.
6. Dutta Soumitra & Stephen Kwan & Arie Segev (1997). Transforming Business in the
Market space: Strategic Marketing and Customer Relationships. CITM.
7. Kennedy, A. Electronic Customer Relationship Management (eCRM): Opportunities and
Challenges in a Digital World
8. Madani.F., Bazini.E., (2013). The role of Insurance companies in albanian economy Journal of
Academic Research in Economics. Volume 5, Number 1, June 2013, 147-161. ISSN: 2066- 0855.
9. Mansell, R. & W. E. Steinmueller (2000). Mobilizing the Information Society: Strategies for
Growth and Opportunity. Oxford University Press.
10. Springer Verlag and Wien (1999). Information Technology and Tourism – A Challenging
Relationship. USA
225
Some Principles for Banks’ Internal Control System in
Albania
Artur Ribaj
Abstract. Internal control involves everything that controls risks to a bank. The objectives of internal
control as a system relate to the reliability of financial reporting, timely feedback on the achievement
of operational or strategic goals, and compliance with laws and regulations. The objectives of internal
control at a specific transaction level refer to the actions taken to achieve the target within the allowed
limit of risk. An effective internal control system reduces process variation, leading to more
predictable outcomes. There are some important documents for regulating the internal control system
as such: The Directive 2006/43/EC “On statutory audits of annual accounts and consolidated
accounts”, published by the European Commission in 2006; 8th Company Law Directive of the
European Union, Article 41; “The internal audit function in banks,” Basel Committee on Banking
Supervision, 2012; “Corporate Governance Principles for Banks” Basel Committee on Banking
Supervision, 2015.
It should be noted that Albania, an EU candidate country, is approaching the EU directives and Basel
principles via legislation and regulation framework. Related to bank’s internal control system, its
structure and responsibilities, Albania has taken periodically the way of accommodating the
requirements of EU directives and Basel principles within the law no. 9662, dated 18.12.2006 “On
Banks on the Republic of Albania”, amended; regulation no. 63, dated 14.11.2012 and lately the new
regulation, dated on 02.09.2015 “On Internal Control System”. This last one, has given the minimum
requirements for setting up an effective internal control system and supporting arrangements by the
three lines of defence model; a strong internal control system, including an independent and effective
internal audit functions, as part of sound corporate governance; the findings of internal audit to be
followed and bank’s management to take appropriate and timely corrective action in response to
internal control weaknesses; the internal audit to provide vital assurance to bank’s board of directors
and supervisors with whom should be an enhanced two-way communication for discussing the risk
areas identified, measures received, etc. This paper aims to inform bankers, researchers, and other
stakeholders with some principles of internal control system regulated in Albania as well as raising
some other related issues to be considerate for ensuring the efficiency and effectiveness of internal
control system of banks in Albanian.
Keywords: internal control system; internal audit function; governance; lines of defence; risk
management.
1. Introduction
Internal control has existed from ancient times. In Hellenistic Egypt there was a dual administration,
with one set of bureaucrats charged with collecting taxes and another with supervising them. As
defined in accounting and auditing is a process for assuring achievement of an organization's
objectives in operational effectiveness and efficiency, reliable financial reporting, and compliance
with laws, regulations and policies. Internal control involves everything that controls risks to a bank
or its subsidiaries. The objectives of internal control as a system relate to the reliability of financial
reporting, timely feedback on the achievement of operational or strategic goals, and compliance with
laws and regulations. The objectives of internal control at a specific transaction level refer to the
actions taken to achieve the target within the allowed limit of risk. An effective internal control system
reduces process variation, leading to more predictable outcomes. Internal control is a key element of
the Foreign Corrupt Practices Act (FCPA) of 1977 and the Sarbanes–Oxley Act of 2002, which
required improvements in internal control in United States public corporations.
The Directive 2006/43/EC “On statutory audits of annual accounts and consolidated accounts”,
published by the European Commission in 2006, set audit committees on the path to becoming a key
226
feature of the corporate governance framework of all EU Member States. The role of the audit
committee, as required by the Directive, was broadly consistent with that set out in many longestablished corporate governance codes, namely to: … monitor the effectiveness of the company's
internal control and risk management systems; monitor the effectiveness of the company's internal
audit function; etc.
According to 8th Company Law Directive of the European Union, Article 41 assigns a duty for
monitoring the effectiveness of risk management and control systems by each “public-interest entity”,
which shall have an audit committee as well. Each respective State Authority shall determine the
composition of audit committees being composed of non-board and board members of the audited
entity. At least one member of the audit committee shall be independent and shall have competence
in accounting and/or auditing. The definition of “public interest entity” varies across the EU, but
generally one of them is considered the activity of banks [1].
On June 28, 2012, the Basel Committee on Banking Supervision (Basel Committee) issued revised
supervisory guidance for assessing the effectiveness of internal audit functions in banks, entitled “The
internal audit function in banks” replacing the 2001 document “Internal audit in banks and the
supervisor’s relationship with auditors”, taking into account developments in supervisory practices
and lessons from the last financial crisis, during which internal audit functions were criticized for their
role in the failure to highlight weaknesses in overall risk management. The revised document
specifically builds on the Basel Committee’s “Principles for enhancing corporate governance”, and
presents 20 principles. These principles are agreed internationally by banking supervisors, following
consultations with the banking sectors. Many of the above principles have also been addressed by
other regulatory and/or association bodies, including but not limited to the United States Commodity
Futures Trading Commission, the U.S. Dodd-Frank Act, and around 90% of corporate governance
codes of the EU member states recognize internal audit as an essential part of the corporate
governance framework.
On July, 2015, the Basel Committee published “Corporate Governance Principles for Banks”. The
revised principles: expand the guidance on the role of the board of directors in overseeing the
implementation of effective risk management systems; emphasize the importance of the board's
collective competence as well as the obligation of individual board members to dedicate sufficient
time to their mandates and to keep abreast of developments in banking; strengthen the guidance on
risk governance, including the risk management roles played by business units, risk management
teams, and internal audit and control functions (the three lines of defence), as well as underline the
importance of a sound risk culture to drive risk management within a bank; provide guidance for bank
supervisors in evaluating the processes used by banks to select board members and senior
management; and recognize that compensation systems form a key component of the governance and
incentive structure through which the board and senior management of a bank convey acceptable risktaking behavior and reinforce the bank's operating and risk culture.
It should be noted that Albania, an EU candidate country, is approaching the EU directives and Basel
principles via legislation and regulation framework. Related to bank’s internal control system, its
structure and responsibilities, Albania has taken periodically the way of accommodating the
requirements of EU directives and Basel principles within the law no. 9662, dated 18.12.2006 “On
Banks on the Republic of Albania”, amended; regulation no. 63 “On core management principles of
banks and branches of foreign banks and criteria on the approval of their administrators”, dated
14.11.2012 and lately the new regulation no. 67, dated on 02.09.2015 “On Internal Control System”.
This last one has given the requirements for setting up an effective internal control system and
supporting arrangements by the three lines of defence model; a strong internal control system,
including an independent and effective internal audit functions, as part of sound corporate
governance; the findings of internal audit to be followed and bank’s management to take appropriate
and timely corrective action in response to internal control weaknesses; the internal audit to provide
vital assurance to bank’s board of directors and supervisors; etc.
2. Some principals of internal control system regulated in Albania
1. In addition to the article 37, point 2 of “Law on Banks” [2], the bank’s board of directors has the
ultimate responsibility for ensuring that executive management/bank’s directorate/bank’s senior
management establishes and maintains an adequate, effective and efficient internal control system,
227
supporting arrangements by the three lines of defence model, and, accordingly, the board should
support the internal audit unit in discharging its duties effectively;
2. In addition to the article 45 and article 46 of “Law on Banks” [3], internal control system and
internal audit unit should have independence, objectivity, professional competence and care in
accordance with best practices and international standards of internal control;
3. The “Three lines of defence” model has to be essential for establishing clearly-structured corporate
governance systems and should provide valid guidance on clear accountability for risk management
and internal control. The “Three lines of Defence” structure is a valid conceptual definition of control
levels: line controls, second-level monitoring controls and third-line independent assurance, but not
to be interpreted as organizational reporting lines. Under the first line of defence, operational
management has ownership, responsibility and accountability for assessing, controlling and
mitigating risks. The second line of controls consists of activities covered by several components of
internal governance (compliance, risk management, quality and other control departments). This line
of defence monitors and facilitates the implementation of effective risk management practices by
operational management and assists the risk owners in reporting adequate risk-related information up
and down the bank. As the third line of defence, an independent internal audit function will, through
a risk-based approach to its work, provide assurance to the bank’s board of directors. This assurance
encompasses all elements of a bank’s risk management framework: from risk identification and
assessment processes to the internal control system as a response to mitigating risks; this includes
communication throughout the bank and to executive management, audit committee and the board of
directors of risk-related information. The responsibility for internal control does not transfer from one
line of defence to the next line;
4. In addition to other responsibilities as per regulation no.63 on 14.11.2012, the risk management
unit and compliance unit, who are considered among the main units of the second line of defence
should be responsible for overseeing the risk management and compliance risk management in view
of the internal control system. These two units should have the adequate authority for accomplishing
their functions;
5. Internal audit function or unit provides independent assurance to the board of directors and advises
the executive management if it asked, on the quality and effectiveness of a bank’s internal control,
risk management and governance systems and processes, thereby helping the board and senior
management protect the bank and its reputation;
6. In addition to the article 38, point 3 of “Law on Banks” [4], the audit committee should: oversee
the bank’s internal audit system; monitor the effectiveness of the internal audit unit; ensure that the
internal audit unit carries out its responsibilities independently and on the basis of internal control
standards; ensure that the bank’s executive management has established and maintains an adequate
and effective first and second line of defence; may propose the appointment or dismissal of the
employees and head of internal audit unit; ensure a regular open dialogue between the
external/statutory auditor, bank’s executive management, supervisory authority and internal audit;
7. Each bank should have a permanent internal audit function that must be independent of the audited
activities, which requires the internal audit function to have sufficient standing and authority within
the bank, thereby enabling internal auditors to carry out their assignments with objectivity. Internal
audit must have adequate resources available, supporting both efficient and effective audit planning
and management. The internal audit function should be accountable to the board of directors, on all
matters related to the performance of its mandate as described in the internal audit regulatory
framework, which articulates as well the purpose, standing and authority of the internal audit function
within the bank in a manner that promotes an effective internal audit function;
8. The completeness of the mandate of internal audit, the scope, budget and plan of auditing will be
consulted from audit committee and then approved by the board of directors. In addition to other
things, internal auditing should have full, free and unrestricted access to any function, activity or unit
under review. Every bank activity (including the outsourced activities by the engagement of external
experts to perform internal audit activities for supporting the internal audit function) or unit should
fall within the overall scope of the internal audit function. The scope of the internal audit function’s
activities should ensure adequate coverage of matters within an annual internal audit plan, which can
be part of a multi-year plan;
9. The determination of the frequency of inspection based on the assessment by a risk-based
methodology for each field of activity and / or organizational units, to focus on those activities /
organizational units that pose more risk by allocating human resources more efficiently; as well as
228
extending the maximum cycle of 2 to 3 years for the checks carried out on those fields of activity and
/ or organizational units valued bank with low risk according to the risk-based methodology;
10. The board of directors to set an organizational unit responsible for tracking and reporting the level
of implementation of the recommendations of the supervisory authorities, external/authorized
auditors, control structures of the parent bank, etc.;
11. Every time, every issue raised by internal auditing for bank’s executive management, as soon as
possible should be reported in written to the bank’s board of directors. The board of directors and
audit committee should have meetings with head of internal audit unit or internal auditors without the
presence of management.
12. In order the employees of internal audit unit to independently fulfill their responsibilities, the
hiring, remuneration, and dismissal will be a decision reserved to the bank’s board of directors, who
might get recommendations from audit committee;
13. Integrity, professional competence, including the knowledge and experience of each internal
auditor and of internal auditors collectively, is essential to the effectiveness of the bank’s internal
audit function. Internal auditors must act with integrity and professionalism;
14. The competency of the head of internal audit unit (“fit and proper”), requiring strong leadership
capability as in addition to technical and communication skills; as well as the adequacy of resources,
both human and technical, including diversity of professional competencies;
15. In addition to other responsibilities related to internal control system, the head of the internal audit
unit should be responsible for ensuring that the internal auditors comply with sound auditing standards
and with a relevant code of ethics which but must be adapted clearly to the legal and regulatory
framework in Albania and bank’s characteristics, activities and culture;
16. The enhanced two-way communication between the supervisory authority and the structures of
internal control system. The frequencies of these meetings should be proportionate with the concept
of proportionality, which mean the bank’s size, the nature and risks of its operations and the
complexity of its organization. They have to discuss the risk areas identified by both parties, measures
received, etc.
18. Others…
Conclusions
1. The challenges arising from the economic situation, and changes in regulatory framework, increase
the pressure for banks to adopt a robust governance framework keeping an effective communication
between structures.
2. Risk management, internal control and audit functions provide reliable information channels for
the bank’s board of directors to monitor the effectiveness of risk management and internal control
system. With regard to risk management, the bank’s board and/or the audit committee needs to
receive, at least yearly, a review of the bank’s risks and its internal control system. They need to have
appropriate information to know how the business model is impacted by major risks, and how value
generation could be enhanced by opportunities or reduced by vulnerabilities of external environment
or particular risks inherent to the activity of the bank and what strategies to be chosen for achieving
the bank’s mission.
3. Ultimately, the bank’s board of directors assisted from audit committee has the final responsibility
for internal control and reporting lines, accomplishing that, through delegation of duties, authorities,
check and balance. Audit committee oversight must rely on an all-encompassing, comprehensive
structure that incorporates all elements of corporate governance.
4. Independence of the audit is vital for the bank life. But, internal audit, external audit and audit
committee are not truly independent because they are employed by the bank (the first one directly,
the second and third under contract by bank’s board of directors). The important feature about audit’s
independence for these three bodies is that it is independent of bank’s executive management and can
therefore give objective assurance to the bank’s board of directors that has the ultimate responsibility
and accountability for ensuring the efficiency and effectiveness of internal control system.
5. The critical issue facing the bank’s board of directors and audit committee is the asymmetry of
information reported to them. It might either overwhelm or disorient judgment of them.
229
Recommendations
1. Rules are meaningless if the bank’s culture is of non-compliance, for that fact, the board and
executive management must set the “tone at the top” of the bank for a corporate culture, which
acknowledges and maintains an effective control environment, where every person within the bank
should be under monitoring by internal controls.
2. The bank’s culture, code of conduct, human resources policies and performance reward systems
are critical components of the internal control system. The culture of the bank should encourage
individuals to report suspected breaches of law or regulations or other improprieties. Critical for this
system to work is the protection of the people who “blow the whistle”.
3. The regulation no. 63, dated 14.11.2012, needs further enhancement adoption with corporate
governance principles for banks, published by Basel Committee on July, 2015 and recommendations
of 8th EU Company Law Directive, Article 41, for the composition of audit committee to be composed
of non-board and board members of the bank with depth background in audit, and at least one member
of the audit committee shall be independent.
4. The bank’s board of directors and audit committee need to know the critical risk issues that require
their attention and to state clearly what information need, its relevance, the format and timing of such
information, for avoiding the asymmetry of information.
5. The bank’s audit committee has to evaluate the reports from these multiple sources and advice the
bank’s board of directors to determine the direction the bank should take. Also, the audit committee
has to ensure board of directors that executive management is taking necessary corrective actions
within an appropriate time frame to address the findings, recommendations, control weaknesses, noncompliances with legal and regulatory framework and other problems identified by internal auditors,
external auditors and supervisory authorities.
6. The audit committee has to have a charter that clearly sets out its role, responsibilities and
accountabilities in providing internal control governance to effectively discharge the requirements
delegated by the board of directors. The audit committee ongoing, but not less than three times per
year, has to report to the board of directors on the status of the bank’s internal control system.
7. The supervisory authorities has to enhance frequently the two-way periodic communication with
bank’s structures of internal control system discussing about risk areas identified by both parties,
understanding the risk mitigation measures taken by the bank, monitoring the bank’s response to
weaknesses identified and supervising if bank management is too strongly influenced by rewards,
such as bonus incentives, and the fear of shareholder demands to ignore or take risks that may lead to
regulatory intervention or, even worse, financial failure.
8. Internal audit function should have the right to attend or observe executive committee meetings
and to have timely access to relevant management information and committee materials.
9. Internal audit function should ensure that its internal audit reports are provided to the board of
directors or the audit committee without bank’s executive management filtering.
10. Banks in Albanian, have to comply as a minimum with the requirements of law no. 9662, dated
18.12.2006 “On Banks on the Republic of Albania”, amended; the regulation no.63, dated 14.11.2012
“On core management principles of banks and branches of foreign banks and criteria on the approval
of their administrators”; and the Regulation no. 67, dated 02, 2015 “On Internal Control System”, and
furthermore illuminating their internal regulatory acts with other Basel-based principles governing
effectiveness, scope and governance.
References:
1.
Law no. 9662, dated 18.12.2006 “On Banks on the Republic of Albania”, amended.
Respectively: Article 37 “Powers of the Supervisory/Steering Council/Board”, Point 2;
Article 38 “Control Committee”; Article 45 “Internal control system”; Article 46 “Internal
control unit”; Article 47 “Financial reports”, Point 3; Article 49 “Audit from the
statutory/external auditor”, point 3; Article 63 “Large exposures”, point 3; Article 76 “Order
230
for the abatement of the unlawful acts”, point 1; Article 80 “Serious breaches”, Point 1;
Article 88 “Submission of Data”, Point 1; Article 89 “Penalizing measures”, Point 3.
2. Regulation “On the Internal Control System” approved by SC of BoA with decision no.67
on 02 September 2015;
3. Regulation no. 63, dated 14.11.2012 “On core management principles of banks and
branches of foreign banks and criteria on the approval of their administrators”
4. “The internal audit function in banks,” Basel Committee on Banking Supervision, 2012;
5. “Internal audit in banks and the supervisor’s relationship with auditors,” Basel Committee
on Banking Supervision, 2001;
6. “Principles for enhancing corporate governance,” Basel Committee on Banking
Supervision, 2010
7. 8th Company Law Directive of the European Union
8. Corporate Governance Principles for Banks, published by Basel Committee on July, 2015
9. Directive 2006/43/EC “On statutory audits of annual accounts and consolidated accounts”,
published by the European Commission in 2006
10. Publications on: Policy position paper on risk management and internal audit; and briefing
on whistleblowing
11. Publications of CIA-Certified Internal Auditor, CIIA – Chartered Internal Auditor, CFSACertified Financial Services Auditor, CGAP-Certified Government Auditor, CCSACertified, Control Self-Assessment, CRMA-Certified Risk Management Auditor and
COSO-Committee of Sponsoring Organizations of the Tread way Commission.
Citations
[1] The EC 8th Company Law Directive defines a public interest entity as: “…entities governed by
the law of a Member State whose transferable securities are admitted to trading on a regulated market
of any Member State within the meaning of Article ..., credit institutions within the meaning of Article
... and insurance undertakings as defined in Article .... Member States may also designate other entities
as public interest entities, for instance entities that are of significant public relevance because of the
nature of their business, their size or the number of their employees…”
[2] The article 37, point 2 of “Law on Banks”: “…The main responsibilities of the Steering Council
shall include amongst others:…taking the necessary and adequate measures for ensuring the integrity
of the financial and accounting system of the bank, including the independent controlling of the bank,
and ensuring the existence of the appropriate systems of control especially with regard to risk
management, operational and financial system of the bank, as well as ensuring compliance with the
law and best practices in the banking system; monitoring and supervising implementation of legal and
regulatory requirements and of the best practices in banking system;…”
[3] The article 45 of “Law on Banks”: “…bank shall organize an internal control system for the
purposes of monitoring the implementation of internal policies and procedures, evaluation of
effectiveness of banking activity and monitoring compliance with law and by-laws; to identify the
exposure of the bank to the type of risks, as well as measuring, administering and monitoring their
levels;...”, and the article 46 of “Law on Banks”: “…internal control unit, as part of the internal
control system with the responsibility for the provision of an effective control over the implementation
of policies, regulations and procedures approved by the bank’s board of directors, as well as provision
of accuracy of information and effectiveness of measures preventing risks faced by the bank…”
[4] The article 38, point 3 of “Law on Banks”: “…controls and supervises accounting procedures and
internal control of the bank, including the procedures defined by the Bank of Albania, and supervises
the implementation of these procedures as well as controls the bank accounts and respective
registrations; considers internal control reports and monitors the way conclusions from such reports
are dealt with; proposes the external auditor and realizes the communication between him and the
internal control of the bank; evaluates the financial situation of the bank based on the report of the
external auditor; controls compliance of the activity of the bank with laws and by-laws and notifies
the Steering Council of the bank about the conclusions; gives an opinion to the Steering Council of
the bank in relation to issues, for which the latter has requested such an opinion; approves the financial
reports and statements prepared by the bank and which the bank intends to publish;…”
231
Katalogimi në botim – (CIP)
Biblioteka Kombëtare e Kosovës “Pjetër Bogdani”
004(496.5)(063)”2015”
569.2(496.5)(063)”2015”
Proceedings of the 4th UBT Annual International Conference on
Business, Technology and Innovation : Computer Science and Communication
Engineering : Information Systems and Security / Editor Edmond Hajrizi. Prishtinë : UBT, 2015. – 232 f. : ilustr. ; 30 cm.
1.Hajrizi, Edmond
ISBN 978-9951-550-14-7

Book of Proceedings - UBT

Transcription

Similar documents

Sunday, November 10, 2013 - Temple Of Deliverance HCG

Internet of Things. A recent visit to CES:

November Newsletter - J. Douglas Hodgson Elementary School

Power devices - Crosslight Software

Calgary Sun Article - Western Lights Artists Group

Internet of Things Security Implications v1.4

Order of Celebration and Praise - Community Church of Lambertville

November newsletter 2

PAPA•RAzzi PAtROl

Order of Celebration and Praise - Community Church of Lambertville