Predicting Short-term Demand and Exploring Influencing Factors in Bike Sharing Systems

Abstract

Bike share systems worldwide are increasingly recognized for their environmental benefits and potential role in promoting active transportation. These systems are particularly valued for addressing the first and last mile problem by providing access to public transport services and facilitating multimodal travel. This study explores the impact of subway and bus operations on the usage of New York City's Citi Bike system. Utilizing a Random Effects Spatial Autoregressive (RE-SAR) model on the counts of bikes rented and returned at each station, to account for spatial and temporal dynamics between bike station usage and public transportation is investigated, with a particular focus on subway and bus operations. The estimation results indicate a positive correlation between both subway and bus arrivals and the usage of the bike share system, with subway access exerting a more significant impact. Furthermore, the study employs a Graph Attention Network (GAT) model, enriched with key factors identified through the RE-SAR model, to forecast short-term usage for bike shares at individual stations. This research underscores the influence of public transit on bike share systems, demonstrating how the environment surrounding the stations and temporal variables affect usage. Additionally, integrating the GAT model into public transit and bike share program applications could significantly enhance user experiences by addressing common issues such as empty or full bike stations, thereby promoting the adoption of this sustainable mode of transportation.