Performance Evaluation of Short-term Travel Time Prediction Model on urban arterials

The capability to predict the short-term travel time on urban major arterials with reasonable accuracy is one of the key functions of an effective Advanced Traveler Information System and Traffic Management System. This paper focused on developing short-term travel time predication models for the major arterials in Beijing City. Six models were developed and tested for the arterial short-term travel time predication problem. The main objective of this study is to investigate the performance of different models so as to identify the operational settings and the anticipated predication accuracy of the models. The real-time data collected from the number plate identified system with video camera on arterials in Beijing. Various scenarios were tested, including 5 min predication, different congestion level, short link versus long link, path-based versus link-based predication, single model versus combined model, to investigate the performance of these models. Mean absolute percentage error was used as a major measure of performance.