A bike-sharing hybrid deep learning model: the London and New York cases

Bike-sharing systems offer a convenient, efficient, and sustainable option for urban transportation, helping alleviate congestion. The growing demand necessitates continued understanding of how the systems are used. Worldwide, overall usage rose in 2022, indicating a shift in bike-sharing habits post-COVID-19. For current and future systems to seamlessly integrate into the urban transport network, achieving precise travel-demand forecasts at a detailed spatiotemporal level is paramount. This study proposes a hybrid deep learning model to forecast bike demand at the dock-station level and to perform network analysis to rank the importance and interactions among docking stations. The model, using historical usage data and weather as the primary drivers, demonstrated effective predictive capabilities for hourly demand at docking stations where specific user types, such as commuters, dominate. The network analysis identified significant docking stations, often located near workplaces, transport hubs, and recreational areas. These insights bolster sustainable transport strategies and forecasting for bike-sharing systems.