A Self-learning algorithm for predicting bus arrival time based on historical data model

The provision of timely and accurate bus arrive time information is very important. It helps to attract additional ridership and increase the satisfaction of transit users. In this paper, a self-learning prediction algorithm is proposed based on historical data model. Locations and speeds of the bus are periodically obtained from GPS senor installed on the bus and stored in database. Historical travel time in all road sections is collected. These historical data are trained using BP neural network to predict the average speed and arrival time of the road sections. Experimental results indicate that the proposed algorithm achieves outstanding prediction accuracy compared with general solutions based on historical travel time.