Bus Arrival Time Prediction with LSTM Neural Network

摘要

Arrival time is a key aspect of passenger information systems. Provision of accurate bus arrival information is essential for delivering an attractive service and necessary to passengers for reducing their waiting time and bus stops and choosing alternative routes. Recently, the same information is used in smart-phone trip planners. In this paper, we explore an LSTM neural network model for bus arrival time prediction. We take into account heterogeneous information about the transport situation, directly or indirectly affecting the prediction travel time. We evaluate the proposed models with bus operation data from Samara, Russia. Evaluation results show that the proposed model outperforms some typical prediction algorithms.