v - UNIPA, DICGIM
Transcription
v - UNIPA, DICGIM
Smart Transport for Sustainable City Francesco Marcelloni Dipartimento di Ingegneria dell’Informazione University of Pisa, Italy E-mail: francesco.marcelloni@unipi.it Alessio Bechini, Beatrice Lazzerini PerLab Projects PerLab “SMARTY” (SMArt Transport for sustainable citY) project funded by “Programma Operativo Regionale (POR) 2007-2013” objective “Competitività regionale e occupazione” of the Tuscany Region” Urban Sensing Social Sensing Analysis of GPS traces “Metodologie e Tecnologie per lo Sviluppo di Servizi Informatici Innovativi per le Smart Cities” project funded by “Progetti di Ricerca di Ateneo - PRA 2015” of the University of Pisa GPS traces similarity Francesco Marcelloni The Smarty Project PerLab “SMARTY - SMArt Transport for sustainable citY”, funded by the Tuscany Region in the framework of Bando Unico R&S - 2012 Francesco Marcelloni The Smarty Project PerLab “SMARTY - SMArt Transport for sustainable citY”, funded by the Tuscany Region in the framework of Bando Unico R&S - 2012 Francesco Marcelloni Our role in the Smarty project PerLab • Urban Sensing • Cooperative air quality monitoring based on low-cost sensors (uSense) • • Privately owned by citizens Deployed in places where they live and spend most of their time • Low-cost system for smart urban parking • Social Sensing • Real-Time Detection of Traffic Congestions from Twitter Stream Analysis • Critical event detection from Facebook events analysis • GPS trace analysis • Real time traffic analysis • Real-time detection of incidents Francesco Marcelloni Social Sensing PerLab • Tweet analysis aimed at • Detecting traffic congestion • Detecting if traffic congestion is caused by an external event • • • • • Soccer match Procession Demonstration Flash-mob … • Notifying (in real-time) users about traffic congestion E. D'Andrea, P. Ducange, B. Lazzerini, F. Marcelloni, «Real-Time Detection of Traffic From Twitter Stream Analysis», IEEE Transactions on Intelligent Transportation Systems , Vol.16, no.4, pp.2269-2283, Aug. 2015. Francesco Marcelloni Social Sensing PerLab • Tweet analysis aimed at • detecting road traffic congestions and accidents • discriminating traffic event due to an external cause (football match, procession, demonstration, flash-mob, etc.) • notifying (in real-time) the users of the traffic event • Facebook event analysis aimed at • monitoring the number of partecipants along the time • notifying the users when the event is likely to be critical Francesco Marcelloni Traffic detection from Tweet analysis PerLab ...I'mstuck stuck in a 7km km ...I'm ...I'm stuck inina a7 7km queue... queue... queue... C Tokenization I7km C ...I'mstuck stuckininaF km Ia7km ...I'm I7C F ...I'm stuck inF a F F queue... A queue... RAF queue... SUM: Status Update Message TRA TTR Stop-word filtering Fetch of SUMs and Pre-processing Stemming Classification of SUMs Stem filtering Feature representation Elaboration of SUMs Text mining elaboration on a sample tweet Text of a sample tweet Sono bloccato in una coda di 7 km... il traffico è incredibile stasera! Voglio tornare a CASA!!! English translation: I'm stuck in a 7 km queue... traffic is unbelievable this night! Wanna get HOME!!! Feature representation [arriv, blocc, caos, cod, km, ..., ..., stasera, traffic, vers, vial]F wq=ln(Ntr/Nq) Tokenization Stop-word filtering tokens <sono>, <bloccato> <in>, <una>, <coda>, <di>, <7>, <km>, <il>, <traffico>, <è>, <incredibile>, <stasera>, <voglio>, <tornare>, <a>, <casa> <sono>, <bloccato> <in>, <una>, <coda>, <di>, <7>, <km>, <il>, <traffico>, <è>, <incredibile>, <stasera>, <voglio>, <tornare>, <a>, <casa> Stem filtering <blocc>, <cod>, <7>, <km>, <traffic>, <incredibil>, <stasera>, <vogl>, <torn>, <cas> Stemming <bloccato> , <coda>, <7>, <km>, <traffico>, <incredibile>, <stasera>, <voglio>, <tornare>, <casa> F relevant stems selected in the learning phase [0, wblocc, 0, wcod, wkm, ..., wstasera, wtraffic , 0, 0]F Francesco Marcelloni [arriv, blocc, caos, cod, km,..., stasera, traffic, vers, vial]F <blocc>, <cod>, <7>, <km>, stems <traffic>, <incredibil>, <stasera>, <vogl>, <torn>, <cas> Traffic detection from Tweet analysis PerLab • Binary classification problem • traffic vs. non-traffic tweets • balanced 2-class dataset of 1330 tweets • best accuracy: 95.75% using an Support Vector Machine (SVM) classifier Prec TP TP FP Rec TP TP FN F -score 1 2 Francesco Marcelloni Prec Rec 2 Prec Rec Traffic detection from Tweet analysis PerLab • Multi-class classification problem • • • • • traffic due to external event vs. traffic congestion or crash vs. non-traffic balanced 3-class dataset of 999 tweets best accuracy: 88.89% using an SVM classifier Francesco Marcelloni Traffic detection from Tweet analysis PerLab • Real-time detection of traffic events • monitoring campaign of areas of the Italian road network • 70 traffic events detected during September and early October 2014 • comparison with official Traffic News Channels • • Autostrade per l’Italia CCISS Viaggiare informati 4 traffic events detected on September, 26th, 2014 • 2 late detection events • 2 early detection events Francesco Marcelloni Facebook event analysis PerLab • Real-time monitoring of events using Facebook • Critical event: at least Ke persons probably will attend the event • Ke is determined based on the event features and context IF 2/3 * Num. Sure + 1/3 Probable > Ke THEN the event is critical • Analysis on the trend of the possible attendees {"type":"EventoFacebookCritico","eventoFb":{ "idFb":"365145446986497", "nome":"open day #master #alta #formazione", "descrizione":"una giornata di incontri ed orientamento per futuri studenti dei nostri master e corsi di alta formazione:\n\n- presentazione delle attività didattiche\n- workshop con coordinatore e docenti \n- incontro con ex alunni\n\nl\u0027\u0027\u0027\u0027open day è aperto a tutti.\n\ninizio corsi novembre 2014\niscrizioni ai corsi entro il 31 ottobre 2014 \n\npossibilità di colloqui individuali. sede: roma. \nper partecipare all\u0027\u0027\u0027\u0027open day è necessario registrarsi online http://goo.gl/rgg8wy", "owner":"75874409682", "location":"accademia di costume e di moda", "startTime":"2014-10-25T11:00:00+0200", "endTime":"data_stimata : 2014-10-25T13:00:00", "pointWKT":"POINT((12.468175254952 41.901237903208)", "partecipazione":{"attending":"22","maybe":"2","declined":"6"}}, "tipoEventoClassificato":"Arte","angleIndex":{"timeInMs":67442172,"estimatedAttending":22}}, Francesco Marcelloni GPS trace analysis PerLab to exploit vehicle GPS traces as traffic sensors Francesco Marcelloni GPS trace analysis PerLab • Spatiotemporal GPS traces analysis aimed at • Detect road traffic congestions and accidents • Notify the users of a traffic alert containing • Affected area • Critical traffic levels o slowed traffic o very slowed traffic o blocked traffic o incident • Detected velocity of vehicles E. D'Andrea, F. Marcelloni, «Detection of Traffic Congestion and Incidents from GPS Trace Analysis», submitted to an International Journal. Francesco Marcelloni GPS trace analysis PerLab • Approach • Matching of GPS traces on the road segments of the digital map of the city • Development of an expert system for traffic and incident detection GPS traces (latitude, longintude, velocity, timestamp, vehicle id ) Digital map Pre-Processing - establish vehicles travel direction, - perform routing, - match GPS traces on digital map Segment Traffic Classification - assign a traffic label to each segment Traffic Alert Notification - perform a spatiotemporal analysis for traffic and incident detection Traffic alert (magnitude, estimated velocity, congested segments) Francesco Marcelloni GPS trace analysis PerLab Road segment classification based on the velocity of vehicles in traffic states in sj the segment with respect to the traffic code velocity blocked vblock very slowed flowing slowed P2 % × vcode j P1 % × vcode j absent vcode j space alert for incident with queue T=1 S very slowed very slowed blocked absent very slowed very slowed very slowed blocked absent very slowed very slowed very slowed blocked absent T=2 time Spatiotemporal analysis of near classified segments in consecutive time intervals T=3 very slowed Francesco Marcelloni GPS trace analysis PerLab • Experimental results • Used SUMO (Simulation of Urban Mobility) • Simulations of GPS traces of 50000 cars and 48 incidents in Pisa, Italy • using SUMO (Simulation of Urban Mobility) framework • Incidents were correctly detected • Incident detection rate: 91.6% • Average detection time: < 7 minutes • It is also possible to detect the congestion propagation in roads close to the incident Francesco Marcelloni GPS trace similarity PerLab • Objective • To understand how much two GPS traces are similar to each other • Current methods • Exploit the concept of Point of Interest • Not suitable to our aims • New concept of similarity based on closeness of the paths • Applications: car pooling Francesco Marcelloni Data Mining for Big Data PerLab • Analysis of a large amount of data collected from different types of sensors • Data mining algorithms for big data • In particular, accurate and interpretable classification and regression systems. • Speed-up close to linear Francesco Marcelloni Questions? PerLab Francesco Marcelloni Department of Information Engineering University of Pisa, Italy E-mail: francesco.marcelloni@unipi.it Francesco Marcelloni