Non-Linear Impacts and Spatial Variations in Multidimensional Built Environments on E-Shopping Decisions: Evidence from Shanghai

While e-commerce has transformed consumption patterns, online shopping behavior remains influenced by the physical environment. Using Shanghai as a case study, this research applies machine learning, SHAP analysis, and K-Means clustering to examine the nonlinear impacts and spatial variations in the built environment on e-shopping. The findings reveal that: (1) E-shopping expenditure follows a long-tail distribution and displays a concentric spatial pattern, peaking between the Outer and Suburban Rings while decreasing within the Inner Ring and beyond the Suburban Ring. (2) Built-environment factors exhibit non-linear effects, with local shopping potential and transit distance playing dominant roles. Indicators such as store density and delivery facility coverage show inverted U-shaped threshold effects, indicating a shift from complementarity to substitution between offline and online retail. (3) The urban space can be clustered into three sub-district types—traditional residential, single-function, and mixed-use—each with distinct e-shopping patterns and drivers. This research highlights the spatial mechanisms shaping digital consumption, providing empirical evidence for context-specific retail planning in megacities.