Nonlinear and spatially heterogeneous effects of the built environment on shared electric-bike demand: evidence from Yangzhou, China

Shared electric bikes (e-bikes) have become important for short-distance travel due to their convenience and power-assisted advantages. However, the nonlinear and spatially heterogeneous effects of the built environment (BE) on shared e-bike demand remain insufficiently studied. To address this gap, this study uses shared e-bike order data from Yangzhou and integrates automated machine learning (AutoML) with GeoShapley to examine these effects. The results reveal that most BE variables exhibit nonlinear effects, with geographic locations, caters, and life services being the most influential factors. Caters, life services, shopping, and work show positive nonlinear relationships with demand, whereas distance to CBD has a negative effect. Moreover, BE effects vary across spatial grid cells; life services stimulate demand in peripheral expansion areas but may become inhibitory in densely developed historic core areas. These findings highlight the importance of considering nonlinear effects and spatial heterogeneity for refined transport policies and differentiated spatial interventions.