Dynamic Travel Time Estimation Using Regression Trees; Final rept

摘要

This report presents a methodology for travel time estimation by using regression trees. The dissemination of travel time information has become crucial for effective traffic management, especially under congested road conditions. In the absence of collected actual observations on travel time, the vehicle speed can be predicted by using regression trees, which in turn is used as a proxy to estimate the travel time. To maintain stable prediction ability in both free flow conditions and near-capacity flow conditions on freeways, the regression tree model developed for this study includes thirteen explanatory variables, categorized in four variable types: traffic flow, incident related, weather data, and time of day. A total of four characterization standards (outliers, weather, incidents, and weekday/weekend) are used to characterize the daily traffic data sets to determine the best regression tree model(s) to predict a day in certain characterization. The results show that not only do the regression tree models have accurate prediction ability of vehicle speed and promising ability to estimate travel time, but also the regression tree models built upon other characterizations are preferred to predict a certain characterization. The loop-detector data on PORTAL (Portland Oregon Regional Transportation Archive Listing) system, for the I5-I205 loop in Portland, Oregon, is used to demonstrate the applicability of regression trees in this report.