Deep trip generation with graph neural networks for bike sharing system expansion

Bike sharing is emerging globally as an active, convenient, and sustainable mode of transportation. To plan successful bike-sharing systems (BSSs), many cities start from a small-scale pilot gradually expand the system to cover more areas. For station-based BSSs, this means planning new stations based on existing ones over time, which requires prediction number trips generated by these across whole system. Previous studies typically rely relatively simple regression or machine learning models, are limited in capturing complex spatial relationships. Despite growing literature deep methods for travel demand prediction, they mostly developed short-term time series data, assuming no structural changes In study, we focus trip generation problem BSS expansion, propose graph neural network (GNN) approach predicting station-level multi-source urban built environment data. Specifically, it constructs multiple localized graphs centered each target station uses attention mechanisms learn correlation weights between stations. We further illustrate that proposed can be regarded generalized model, indicating commonalities GNNs. The model evaluated realistic experiments using multi-year data New York City, results validate superior performance our compared methods. also demonstrate interpretability uncovering effects features interactions stations, provide strategic guidance location selection capacity planning.