Nonlinear Effects of Land Use Features on Metro Ridership: An Integrated Exploration with Machine Learning Considering Spatial Heterogeneity

Abstract

This study explored nonlinear effects of land use features and developed an analytical framework using integrated gradient boosting decision tree for modeling metro ridership. Station-level boarding and alighting ridership at different times of day was derived from smart card records and used as the dependent variable. The values of 19 independent variables, including land use, were calculated based on the directional and size-various catchment area defined by shared bike’s origin-destination data. This framework considering spatial heterogeneity demonstrated goodness-of-fit and prediction power, which has been ignored in previous studies. Furthermore, this framework provided information for modeling based on geographical weighted regression and global machine learning models. The local relative importance mapping selected land use variables with varying impacts across the Shanghai, which differs from the usual averaging into one value in global machine learning models. Meanwhile, the nonlinear relationship between influencing variables, such as leisure, business, and shopping, demonstrated a positive trend with boarding and alighting ridership and spatio-temporal heterogeneity with the effective range and threshold effect. Rather than focusing on increasing development density to boost metro ridership, this study evaluated the saturation of station-level land use to facilitate accurate decision-making based on station-area planning, and investment priorities of city areas.

Keywords: metro ridership, land use, spatial validation, nonlinear effects, spatial heterogeneity