Does location of entrance/exit matter? Reexamining built environment factors of commuter ridership using high-quality LBS data

Abstract

Previous studies have primarily concentrated on metro ridership and its built environment, while often neglecting the role of entrances or exits in each metro station. This oversight can be attributed to the difficulties of applying large-scale spatial data with sufficient precision to accurately identify ridership associated with single entrances or exits. In this study, we used a new type of Location Based Services (LBS) data, the commuter Origin-Destination (OD) data, to calculate Entrance/Exit Ridership (EER) in Xiamen, China. Using the Ordinary Least Squares (OLS) model, and the Random Forest (RF) model with the Shapley Additive exPlanations (SHAP), we explored how EER and the Proportion of Entrance/Exit Ridership (PEER) are subject to the built environment factors at the global level and the station level. The results indicated that more travelers are distributed in some regions of the island and some business parks, but these stations also demonstrate wide heterogeneity in terms of PEER. The density category variable shows a high positive correlation with EER, and shopping POI and intersection have a negative correlation with EER. Compared with EER, PEER is more relevant to more variables, especially the variables related to entrance/exit characteristics. According to the RF model, the influence of built environment variables on the EER and PEER also shows nonlinear and threshold effects. Our findings show that planners should consider the location and number of entrances/exits combined with built environment, in addition to the distribution of working and resident population. This study aims to improve passenger experience through enhanced travel convenience and comfort, providing a foundation for evidence-based policy development and urban design strategies in transit-oriented development.