Shared dockless e-scooter and metro station built environment: its nonlinear relationship using random forest modeling

Determinants of varying adoption rates of e-scooters across areas remain unclear, particularly in built environments near metro stations. Methodologically, linear regression modelings have limitations in capturing complex relationships between the built environment of metro stations and e-scooter ridership. This research uses random forest modeling to examine the nonlinear relationship between the built environment, focusing on station area and shared dockless e-scooter trips. By emphasizing metro stations as critical nodes for multimodal connectivity, the research identifies built environment characteristics – such as activity density, population density, compact index score, and job-housing balance – significantly influence higher e-scooter usage. Though less significant, street design variables like road and intersection densities show nonlinear effects. This study confirms that development factors in metro station areas are more substantial than street design factors in promoting e-scooter ridership, highlighting the need to optimize transit-oriented development and multimodal transportation strategies near metro stations.