Time-aware trip generation for bike sharing system planning

Abstract

<p>Due to its enormous social benefits, bike sharing is becoming increasingly popular all over the world. To develop a station-based bike sharing system (BSS), many cities start with a small area and gradually expand the BSS network by adding new stations. The successful planning of system expansion relies on accurate prediction of not only the overall trip intensity of new stations but also its temporal distribution, which remains underexplored in the literature. To this end, this study investigates the problem of time-aware trip generation (TTG) for BSS planning, which aims to forecast the number of trips generated by new stations at different time periods. This task, however, is challenging due to the complex spatiotemporal dependencies in bike sharing demand and the lack of historical data. To address these challenges, we propose a multitask graph neural network with temporal embedding (MTG-TE) for TTG. Specifically, a spatiotemporal feature extractor is developed to leverage diverse urban context features for encoding spatial and temporal dependencies using graph neural networks and representation learning techniques, respectively. Furthermore, a multi-task demand predictor is designed to predict the daily trip intensity and its hourly distribution as two distinct tasks. Finally, extensive experiments on real-world data from New York City demonstrate the superior performance of our method compared with existing baselines.<br></p>