Investigating nonlinear effects of built environment factors on the integration of bike-sharing and the metro

Understanding the integration of metro and shared bicycle is essential to solve the last-mile problem and promote multimodal transportation. Although previous studies have investigated the connection of shared bicycles to metro systems, few have explored the nonlinear relationship between the influential factors and the shared bicycle pickup demand near the metro stations. This study adopts a CatBoost model to describe the nonlinear effect of metro ridership, built environment, and sociodemographic characteristics on the number of shared bicycles as a feeder of metro in Beijing, China. The Shapley additive explanations model is used to intuitively visualize the results. We found that CatBoost can enhance the prediction power and obtain the relative importance ranking of variables, among which the first is the distance from the central business district. Furthermore, there exist thresholds for the influential ranges and interaction of extracted features. Transportation planners and operators can refer to the findings of this study for making relevant policies to promote intermodal transit trip and help address traffic congestion issues.