Bus Travel-Time Prediction Based on Deep Spatio-Temporal Model

Bus travel time estimation in urban city is of great importance, which reduces passengers' waiting time and improves the quality of service of bus transportation. However, the travel time estimation is affected by various factors, including spatio-temporal dependencies (e.g. traffic conditions and road networks) and external factors (e.g. weather). Moreover, the bus dwelling and transit time are predominantly affected by different factors and hence have different patterns, with a fact that there are not so much study on how to divide the dwelling and transit areas and to build independent models for them. In this paper, we propose an end-to-end deep learning framework for Bus Travel Time Estimation (called DeepBTTE) where the target path is of arbitrary length. Two independent spatio-temporal components that use 1D-CNN and LSTM are adopted to estimate the dwelling time and transit time separately, which are then combined for the final estimation. We conduct experiments to evaluate our model using a real-world dataset. The experimental results show that our approach significantly outperforms other existing methods.