Mining spatio-temporal data for travel time estimation in urban traffic networks

scientific master's theses

• Title (croatian): Dubinska analiza prostorno-vremenskih podataka za procjenu trajanja putovanja u urbanim prometnim mrežama
• Author: Hrvoje Marković; VIAF: 305633578
• Mentor: Bojana Dalbelo-Bašić (mentor)
• Committee member: Zoran Kalafatić (predsjednik povjerenstva)
• Committee member: Bojana Dalbelo-Bašić (član povjerenstva)
• Committee member: Hrvoje Gold (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; (Department of Electronics, Microelectronics, Computer and Intelligent Systems); Zagreb
• Defense date and country: 2007-03-27, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing; Artificial Intelligence
• Universal decimal classification (UDC): 004 - Computer science and technology. Computing. Data processing

Abstract

This master thesis investigates travel time estimation by mining spatio-temporal data acquired from vehicles equipped with GPS receivers in urban traffic networks. GPS based travel time estimation is a novel approach and urban traffic networks that are examined in this work present more complex networks than motorways or highways. Differences faced when using GPS data in contrast to data collected from roadside sensors are investigated. Topics such as map matching, outlier detection and travel time variability reduction are also covered. Three different models for travel time estimation (travel time table, seasonal autoregressive integrated moving average model and k-nearest neighbour model) are investigated. It has been shown that the model commonly giving most accurate estimation in motorways and highways (such as seasonal ARIMA that is presented in this work) is not the most suitable for travel time estimation with GPS data in urban networks. This work has shown that the most suitable model for GPS data in urban networks is k-nearest neighbour non-parametric regression model. The experimental portion is performed on GPS data collected by vehicles travelling through the road network of city of Zagreb. The model identified as the most suitable one for travel time estimation, k-nearest neighbour non-parametric regression, can easily be integrated in already available fastest-route guidance systems. Since the model gives dynamically dependent travel times, those systems can be upgraded into dynamic fastest-route guidance systems.

Abstract (english)

Ovaj magistarski rad istražuje procjenu trajanja putovanja pomoću dubinske analize prostorno-vremenskih podataka prikupljenih pomoću vozila s GPS uređajima u urbanim prometnim mrežama. Procjena trajanja putovanja pomoću GPS podataka predstavlja novi pristup, a u ovom su radu istražene urbane prometne mreže koje su složenije od međugradskih cesta ili autoputa. Prikazane su razlike koje nastaju kod korištenja GPS podataka za razliku od podataka prikupljenih pomoću senzora postavljenih uz cestu. Također su pokrivene teme kao što su pridruživanje podatka odgovarajućem cestovnom segmentu, otkrivanje ekstremnih vrijednosti i redukcija varijabilnosti vremena trajanja putovanja. Istražena su tri različita modela (tablica trajanja putovanja, sezonski ARIMA model i model k-najbližih susjeda). Također je pokazano kako model koji često daje najtočnije predviđanje kod međugradskih cesta ili autoputa (kao što je sezonski ARIMA model predstavljen u ovome radu) nije odgovarajući za procjenu trajanja putovanja pomoću GPS podataka u urbanim mrežama. Ovaj je rad pokazao kako je neparametarska regresija k-najbližih susjeda odgovarajući model za GPS podatke u urbanim mrežama. Eksperimentalni je dio izveden na GPS podacima prikupljenim pomoću vozila koja su vozila cestama grada Zagreba. Predloženi model za procjenu trajanja putovanja, neparametarsku regresiju k-najbližih susjeda, je moguće jednostavno integrirati u već postojeći sustav za pronalazak najbržeg puta. Kako model daje dinamički ovisna vremena trajanja putovanja, takve je sustave moguće nadograditi u sustave za pronalaženje najbržih putova koji ovise o vremenu polaska.

• Keywords: mining spatio-temporal dana; travel time estimation; urban traffic networks; GPS dana; seasonal ARIMA; k-nearest neighbour non-parametric regression
• Keywords (english): dubinska analiza prostorno-vremenskih podataka; procjena trajanja putovanja; urbane prometne mreže; GPS podaci; sezonski ARIMA model; model k-najbližih susjeda
• Language: english
• URN:NBN: urn:nbn:hr:168:452732
• Study programme: Title: Postgraduate master programe in computing; Study programme type: university; Study level: postgraduate; Academic / professional title: magistar inženjer računarstva (magistar inženjer računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2020-03-19 08:04:15