Predicting Bus Arrival Time Using BP Neural Network and Dynamic Transfer

Abstract Precisely predicting bus arrival time and offering the optimal transfer scheme are the two key factors for realizing the intelligent transportation. Most studies only use historical data to predict arrival time, which results in large errors. In this paper, considering real-time bus operation data and transfer schemes, we propose a fusion scheme to predict bus arrival time and improve prediction accuracy. Firstly, it proposed a based link bus arrival time prediction technology according to its real-time operation data; Secondly, the BP neural network model and correction layer is adopted to predict the bus arrival time when the real-time bus operation data of some interval are missed. Finally, a dynamic transfer scheme based on the shortest time priority is designed according to the prediction of bus arrival time. Experimental results demonstrate the prediction accurate of our proposed scheme.