Identifying potential upgradable bus stop locations with on-demand shuttle ridership with VIA data in Jersey City

Abstract

This study aims to identify transit stop locations that have the potential to be upgraded to serve more mobilities using real-life on-demand shuttle service trip data through a novel machine learning framework. Leveraging existing infrastructure, this approach ensures efficient resource allocation and promotes diverse transportation modes. The correlation between demographic and built environment characteristics and potential transit stops offers valuable insights for urban planners, contributing to sustainable and user-friendly urban mobility. The study’s contributions include using real-world shuttle service data from multiple years and complex sources of built environment features including computer vision to pinpoint transit stops for conversion into mobility hubs, enhancing the urban transportation network by promoting resource efficiency. Additionally, it introduces a framework using eXtreme Gradient Boosting and SHapley Additive exPlanations values to understand multimodal mobility hubs. These insights guide urban planners in designing hubs that improve efficiency, reduce travel time, and alleviate congestion. Findings indicate that successful mobility hubs are characterized by higher building density, diverse land use, and well-connected street networks. Increased building footprint ratios and road densities are associated with higher shuttle and transit usage, highlighting the role of urban density and connectivity. Mixed-use environments with high land use entropy attract more shuttle destinations, emphasizing the importance of integrating residential, commercial, and recreational spaces. Effective mobility hubs are likely found in dense, mixed-use urban areas with excellent street connectivity, where Pick-Ups are frequent in areas with less biking and walking infrastructure, and Drop-Offs concentrate in mixed-use areas.