Zaitun Time Series booklet
Transcription
Zaitun Time Series booklet
Zaitun Time Series Developed by: Zaitun Time Series Developer Team http://www.zaitunsoftware.com Background Today, time series analysis has become of the most important and widely used branches of mathematical statistics. Time series analysis consists of some methods which attempt to understand such time series data, to understand the relationship or the influence of some events, include estimating the strength of influence among events. Time series data is also used to forecast future events based on known past events by using regression line and trend. Time series analysis fields of application range from physics to neurophysiology and it covers such well known areas of economic forecasting, study of biological data, control system, selling forecasting, process and quality control, finance analysis, stock market analysis, and census analysis. A tool to help the analysis of time series is very required. It can help the process of analysis to be easier. A tool which can help the analysis of time series is a time series application which provides a simple way in modeling and forecasting time series data. Today, there are a lot of time series applications which have been developed to help the analysis of time series data. But then, there are a lot of time series applications which are developed for commercial use and expensively sold. Free time series applications are also available, but they are not easy to use. The required time series application is the application which is simple, user friendly, easy to use, understandable, and free. Zaitun Time Series Application Zaitun Time Series is a free application which is designed to help and solve the problem about the time series data. Zaitun Time Series is originally developed by “Time Series” team as the final project of four years diploma in Sekolah Tinggi Ilmu Statistik (STIS), Jakarta, 2007. The team consisted of some student from statistic computation major. Until today, the development of Zaitun Time Series is still being continued by the developer team in zaitunsoftware.com. The first version of Zaitun Time Series (version 0.1.1) was finished at August 2008, and in October 2008 Zaitun Time Series was published and could be downloaded freely at http://www.zaitunsoftware.com. Now, the latest version of Zaitun Time Series is version 0.1.4. Zaitun Time Series is designed to help the analysis of time series data. It provides the easy way in modeling and forecasting time series data. Compared to another time series application, Zaitun Time Series is a time series application which is free, easy to use, user friendly, and can be used for any purposes. It provides several analysis of statistics and neural network, it also provides graphical tool to help the analysis of time series becoming easier. The analysis of statistics and neural network consisted of trend analysis, decomposition, moving average analysis, exponential smoothing, correlogram, and neural network. The graphical tools consisted of time series plot, actual and predicted plot, Actual and Forecasted Plot, Actual vs Predicted Plot, Residual Plot, Residual vs Actual Plot, and Residual vs Predicted Plot. The Uniqueness of Zaitun Time Series 1. Simple and easy to use The interface is designed by considering the ease of user. The interface is consisted of three main views, they are project view, variable view, and result view which simplify the management of the time series data and the analysis or forecasting result. 2. The ease of pre-analysis of the data Variable view is consisted of three views, they are spreadsheet view, graphic view, and statistics view. These help the pre-analysis of the characteristics of a variable through the graphical view and the value of statistics. 3. The ease of understanding the analysis and forecasting result The result of analysis is viewed in two ways; they are table and graphical view which help the user to understand the statistics value in the model. 4. A simple way to do data forecasting Data forecasting can be done in an easy way, only by clicking the Forecasted option and fill the number of data forecasting step. The result of forecasting in graphical view can be showed only by clicking the Actual and Forecasted option (see the picture beside). 5. The model of statistics and neural network are available The model of statistics and neural networks are available to do the analysis and forecasting of time series data. The models are: Trend Analysis. Linear, Quadratic, Cubic, and Exponential Decomposition. Multiplicative and Additive Moving Average. Single Moving Average and Double Moving Average Exponential Smoothing. Single Exponential Smoothing, Double Exponential Smoothing (Brown), Double Exponential Smoothing (Holt), and Triple Exponential Smoothing (Winter) Correlogram. Level, First Difference, Second Difference Neural Networks. 6. External data To connect with the external data through this application is quite easy. This application is facilitated with the import and export data feature which import and export the data from and to excel and CSV file. The Features of Zaitun Time Series 1. Creating a time series data in certain frequency. Zaitun Time Series has facility that helps user to make the new project which will be filled with time series data in certain frequency. The frequencies consists of annual, semi annual, quarterly, monthly, weekly, daily, daily (1 week, 6 days), daily (1 week, 5 days), and the sequence of time series data which has an unidentified frequency. 2. Variable and Group Time series data in Zaitun tme Series is organized into variable and group. The variable in Zaitun Time Series represents a single time series variable. Group represents the collection of time series variables. 3. Spreadsheet View, Graphic View, and Statistics View To simplify the user’s job, user can view the value of variable and group in three views, they are spreadsheet view, graphic view, and statistics view. Spreadsheet view shows the value of variable in a grid which will simplify the process of entry or change the value of variable. Graphic view will show the variable in a graphic line which will simplify the graphical analysis of the component of the time series variable (e.g. trend, cycle, seasonal, and irregular). Statistics view shows the descriptive statistics which simplify the analysis of statistical characteristic of a variable. 4. Time series analysis model The model of statistics and neural network to do the analysis and forecasting of time series data are available, they are: Trend Analysis Zaitun Time Series provides the feature to do the trend component analysis of a time series data. Several type of trends are available, they are linear, quadratic, cubic, and exponential. Decomposition Zaitun Time Series provides the feature to do decomposition analysis of time series data. Moving Average Zaitun Time Series provides the feature to do moving average analysis. Single and double moving average analyses are available. Exponential Smoothing Zaitun Time Series provides the feature to do the analysis of Exponential Smoothing. The supported model of Exponential Smoothing consists of Single Exponential Smoothing, Double Exponential Smoothing (Brown), Double Exponential Smoothing (Holt), and Triple Exponential Smoothing (Winter). Correlogram Zaitun Time Series provides the feature to view the value of Auto Correlation Function (ACF) and Partial Auto Correlation Function (PACF) of time series data, include the graphic. Neural Networks Zaitun Time Series provides the feature to make a neural network model of time series data. 5. Graph There are some graphs which help the user to understand the data and the result of analysis, they are: Time Series Plot. Viewing a line graph of a variable. Actual and Predicted Plot. Viewing line graph for actual and predicted value of a model. Actual and Forecasted Plot. Viewing line graph for actual and forecasted value of a model. Actual vs Predicted Plot. Viewing the scatter plot between actual and predicted value. Residual Plot. Viewing the line graph for residual value of a model. Residual vs Actual Plot. Viewing the scatter plot between residual and actual value. Residual vs Predicted Plot. Viewing scatter plot between residual and predicted value. 6. Variable Transformation Zaitun Time Series provides the facility to transform the variable. Several types of supported transformation are differencing, seasonal differencing, logarithm, and quadratic equation. 7. Import and Export the Data Zaitun Time Series is facilitated with the import and export data feature which are able to import and export the data from and to excel and CSV file. This feature will make the work easier, especially for user with different format of data. The Methodology The methodology of Zaitun Time Series consists of: 1. Computation based on matrices. Most of the computation calculation of Zaitun Time Series is based on matrices. The matrices operations such as summation, multiplication, and determinant are used to count the value of analysis. 2. Numerical computation The method of numerical computation such as the calculation of integral, differential, the method of optimization are used to calculate the value of statistics function, statistics distribution, and the optimization of the value of parameter model. 3. Neural Network Zaitun Time Series uses Encog-CS external library to make a neural network model. Encog is a LGPL licensed library which is used in neural network programming and robot programming. Encog-CS can be downloaded at http://code.google.com/p/encog-cs. 4. Object Oriented Programming Zaitun Time Series is developed by using the object oriented programming method. Every data and function in this method is classified into classes and objects. Each object can receive the message, process the data, and send message to another object. Object oriented programming approach gives more flexibility way to manage the program and the codes. 5. C# Language and .NET Framework 2.0 More than 40.000 lines of C# codes in .NET Framework 2.0 have been written to develop Zaitun Time Series. C# Language and .NET Framework 2.0 have been developed by Microsoft and widely used to develop a lot of application include a scientific application. Awards 1. Indonesia ICT Award 2009, Research and Development Category Zaitun Time Series won the Indonesia ICT Award 2009, Research and Development Category given by Department of Communication and Information, Republic of Indonesia. 2. Softpedia 100% Clean Award Softpedia guarantees that Zaitun Time Series is 100% clean, not containing any malware, spyware, virus, Trojan or backdoor. 3. Free Download Manager User Choice Award This award is achieved from Free Download Manager website since many website visitors choose Zaitun Time Series to be downloaded. Statistics of Website Visitor Here is the statistics of http://www.zaitunsoftware.com visitor based on country since October 2008 until February 2009. The top five website visitor based on original country is counted by the highest hits count. The number of the visitor is obtained from the report of Advanced Web Statistics 6.9 which has been installed in Zaitun Time Series website server located in US. October 2008 No Country 1. United States 2. Hong Kong 3. Spain 4. Germany 5. Indonesia Other Countries Total November 2008 No Country 1. United States 2. Hong Kong 3. Germany 4. Indonesia 5. Brazil Other Countries Total December 2008 No Country 1. United States 2. Hong Kong 3. Australia 4. Canada 5. Indonesia Other Countries Total Hits 16,232 918 439 317 174 1,098 19,178 Hits 10,472 6,727 565 399 296 3,929 22,388 Hits 16,457 2,044 633 542 257 4,721 24,654 January 2009 No Country 1. United States 2. Indonesia 3. Hong Kong 4. Australia 5. Great Britain Other Countries Total February 2009 No Country 1. Thailand 2. United States 3. Indonesia 4. Australia 5. Brazil Other Countries Total March 2009 No Country 1. United States 2. Indonesia 3. Russian Federation 4. Egypt 5. Italy Other Countries Total April 2009 No Country 1. United States 2. Indonesia 3. Russian Federation 4. India 5. Spain Other Countries Total Hits 50,744 965 910 733 678 10,070 64,100 Hits 10,319 7,309 6,015 1,773 1,019 11,757 38,192 Hits 6,126 5,905 1,985 1,287 1,116 16,084 32,503 Hits 5,760 2,934 1,681 876 782 10,307 22,340 May 2009 No Country 1. United States 2. Indonesia 3. Russian Federation 4. Saudi Arabia 5. France Other Countries Total June 2009 No Country 1. United States 2. Indonesia 3. Poland 4. Russian Federation 5. Germany Other Countries Total July 2009 No Country 1. United States 2. Indonesia 3. Poland 4. Russian Federation 5. India Other Countries Total August 2009 No Country 1. Indonesia 2. United States 3. Russian Federation 4. Great Britain 5. Poland Other Countries Total Hits 6,582 4,514 3,733 2,205 993 15,306 33,333 Hits 7,688 5,909 3,208 3,062 901 14,094 34,842 Hits 13,451 10,095 5,275 4,551 1,889 21,385 56,646 Hits 14,493 7,946 3,429 1,329 1,037 12,966 41,200 Users and visitors comments Here are the comments of Zaitun Time Series’ users and site visitors: “…I have to say a very big CONGRATULATIONS to you all in the team for producing such a wonderful useable piece of software free to the world and of course to your teachers and supervisors. You have done a good thing and you may never know all the good that will come of what you have done. I can't tell you how impressed I am. ….” [Joe Kenyon] “…many thanks for the amazing software you had done…” [Ahmed Hamdy] “… I think the software is marvelous, and I can use it frequently in my daily work. Thanks for putting it together….” [Tim Altom] “… I use time series analysis of various metrics for clients. Most are very simple and can be described with regression lines, but many are erratic and require more sophisticated analysis, such as moving average or exponential models. Major packages will do these things, but they are generally very expensive, and the company cannot afford them right now. Others, like R, are powerful, but difficult for others in my company to use. Your software is free, easy to use, and very fast. It helps me a lot. I also teach statistics to college students, and I have recommended Zaitun Time Series to others on campus…” [Tim Altom] “Curretly i'm using the software and it's doing well especially in my job” [Anonymous] “…Really nice software…” [André Carlucci] “Your package is unusual and takes a bit of getting used to. Nevertheless I am grateful for it.” [Mike London] “….I dont expect that u guys are so cool …” [Azzy Brastorm] “…good job guys…” [Hanung Pramusito] Next Development 1. Statistics Model Adding more complex statistics models and statistics tests to do the analysis and forecast time series data. Some models which can be added into Zaitun Time Series: - Auto Regressive Integrated Moving Average (ARIMA) - Auto Regressive Conditional Heteroscedasticity (ARCH) - Generalized Auto Regressive Conditional Heteroscedasticity (GARCH) - Vector Auto Regression (VAR) - Unit Root Test - Co-integration Test 2. Neural Network Model Expanding the Neural Network Model in Zaitun Time Series - which is Fully Connected Feed Forward Neural Network with Back Propagation Training to increase the performance of data forecasting. It can be done by using the other topology of neural network such as Radial Basis Network, and Non Fully Connected Neural Network, using another training algorithm such as QuickProp, RPROP, and Lavenberg Marquardt, adding the other method of Artificial Intelligence such as genetic algorithm or fuzzy logic. 3. Automatic Forecasting Statistics models and neural network model need the involvement of user in determining the best model which will be used to do the data forecasting. The users should determine the certain parameter until they can obtain a good result of forecasting. The limitation can be solved by developing an intelligent technique to do an automatic forecasting. The intelligent techniques such as expert system, fuzzy logic, and the genetic algorithm can be combined with the current model in determining the best value of parameters of the model which is able to set a good result of forecasting. 4. The connection to the external database The connection to the external database is very useful for user who saves his data inside a certain database application. The facility to import and export from and to the database will simplify the job to do the data analysis. 5. Live Stock Market Today, some sites provide data of the fluctuation of the current stock market, such as Google Finance and Yahoo Finance. The facility to import the data from these sites will simplify the user who wants to see the fluctuation of stock market. It also helps to do the analysis of time series of the data include to forecast the data by using the model which is available in the application. The result of forecasting can be used to help the process of stock exchange transaction. Zaitun Time Series v 0.2.1 Release Plan Some features which will be included in Zaitun Time Series v 0.2.1: 1. Multiple Regression Analysis 2. New Data Type to help visualization of sock market and forex data 3. New Import Feature to acquire live stock market data and forex data from online data provider like Google Finance or Yahoo Finance The Developer Team Rizal Zaini Ahmad Fathony Core Development, Programmer Suryono Hadi Wibowo Interface Designer, Programmer Lia Amelia Website, Documentation Past Developers: Almaratul Sholihah, Muhamad Fuad Hasan, Rismawaty, Wawan Kurniawan, Aris Wijayanto, Dewi Andriyanti.