Prediction of Bus Travel Time Over Unstable Intervals between Two Adjacent Bus Stops

摘要

This paper addresses the problem of predicting bus travel time over unstable intervals between two adjacent bus stops using two typesof machine learning techniques: ANN and SVRmethods. Ourmodel considers the variability of travel time because the travel time isoften influenced by stochastic factors, which increase the variance of travel time over an interval between inter-time periods. Thefactors also affect the variance of the travel time over the interval at the same time period between inter-days. In addition, the factorsshow some correlation of travel time over the interval between time periods in a day. The performance of the proposed model isvalidated with real bus probe data collected from November 21st to December 20th, 2013, provided by Nishitetsu Bus Company,Fukuoka, Japan. We demonstrated the impact of two types of input variables for the prediction in off- and on-peak (rush hour)periods. The results show that the two types of inputs can effectively improve the prediction accuracy. Moreover, we compared theproposed method with our previous methods. The experimental results show the validity of our proposed method.