Detecting and studying non-recurring congestion in Helsinki : A study of non-recurring congestion detection methods
Visualitza/Obre
master-thesis-traffic-patterns-v6.pdf (2,837Mb) (Accés restringit)
Estadístiques de LA Referencia / Recolecta
Inclou dades d'ús des de 2022
Cita com:
hdl:2117/356368
Realitzat a/ambChalmers tekniska högskola
Tipus de documentProjecte Final de Màster Oficial
Data2021-10-11
Condicions d'accésAccés restringit per decisió de l'autor
Tots els drets reservats. Aquesta obra està protegida pels drets de propietat intel·lectual i
industrial corresponents. Sense perjudici de les exempcions legals existents, queda prohibida la seva
reproducció, distribució, comunicació pública o transformació sense l'autorització del titular dels drets
Abstract
The traffic related data collected from sensors on the road has been growing tremendously. This trend requires new tools to process and analyse the records to extract useful information such as incident detection. For this reason, this thesis explores three different methods extracted from literature to detect road incidents (non-recurring congestion) in Helsinki and analyse them using delay as a traffic congestion metric with one year of data records (i.e., 2018). The three methods tested are based on: dynamic threshold, delay prediction, and outlier detection. The first method uses percentile score to set the thresholds that vary with time, whereas the last two methods are performed employing the algorithm k-Nearest Neighbours (k-NN). The methods validation is tested using a data source that contains part of the incidents occurred during the recordings. This data source is uncompleted as the major part of the incidents are usually not reported. The results shows that k-NN outlier detection outperform the other two methods as reach the 60% of accuracy with the most reasonable number of non-reported incidents (4,35%). Therefore, the research expose that outlier methods are one of the easiest ways to detect incidents in traffic data. Such methods trend to be simpler and obtain better results than more complex techniques. Finally, the incident analysis shows that the most severe accidents occur during the morning rush hour, whereas the afternoon rush hour concentrates the major number of incidents
MatèriesTraffic accidents -- Mathematical models, Traffic safety -- Helsinki (Finland) -- Software, Circulació -- Accidents -- Models matemàtics, Seguretat viària -- Hèlsinki (Finlàndia) -- Programari
TitulacióMÀSTER UNIVERSITARI EN ENGINYERIA INDUSTRIAL (Pla 2014)
Localització
Col·leccions
Fitxers | Descripció | Mida | Format | Visualitza |
---|---|---|---|---|
master-thesis-traffic-patterns-v6.pdf | 2,837Mb | Accés restringit |