Data Warehouse Based Traffic Jam Detection
Authors
Eliasen, Jan ; Kjær, Casper ; Urban, Helen
Term
4. term
Education
Publication year
2002
Abstract
Når bilister advares om trafikkøer i tide, kan de vælge en anden rute. Dette speciale designer og evaluerer tre metoder til automatisk at opdage trafikkøer. For at understøtte testene blev der opbygget et datalager (en stor database med historiske trafikoplysninger) for at undersøge, om det kan identificere ikke-periodiske, uregelmæssige køer. En prototype blev implementeret, og de tre metoder blev testet på realistiske, simulerede trafikdata fra en datagenerator. Metoderne blev sammenlignet indbyrdes og også anvendt i kombination. Forsøgene tyder på, at trafikkøer kan opdages, og at resultaterne bliver bedre, når centrale parametre justeres. Den vigtigste styrke ved datalagre, præaggregering (dvs. at gemme sammenfattede data på forhånd), blev dog ikke udnyttet her, så en almindelig relationel datamodel kunne i stedet have været brugt.
Early warnings about traffic jams let drivers choose alternative routes. This thesis designs and evaluates three methods to automatically detect traffic jams. To support testing, a data warehouse (a large database of historical traffic information) was built to examine whether it can identify non-periodic, irregular traffic jams. A prototype was implemented, and the three methods were tested on realistic, simulated traffic data produced by a data generator. The methods were compared with each other and also tried in combination. The experiments indicate that traffic jams can be detected, and that performance improves when key parameters are tuned. The main advantage of data warehouses, pre-aggregation (i.e., storing summarized data in advance), was not used here, so a standard relational database model could have been used instead.
[This abstract was generated with the help of AI]
Documents