Travel Time of Buses Based on GPS Trajectory Data: Analysis and Prediction

Travel time of buses from start station to end station is a key indicator of operational performance. It is of significance for arranging and dispatching vehicles by operators. In this paper, we analyze the travel time of buses with GPS (global positioning system) trajectory data. The results show that there exist obvious peaks during rush hours on weekdays, which means the travel time is larger in rush hours than other hours. Moreover, the distribution of travel time follows a right skewed distribution. Additionally, we divide the travel time into running time of sections and dwell time at stations and analyze their performance specifically. Finally, a BP neural network prediction model is introduced to predict travel time of buses using the historic travel time with time intervals. The results show that it outperforms Gauss fit and Kalman filter methods.