Exploring Spatial Heterogeneity of E-scooter’s Relationship with Ridesourcing using Explainable Machine Learning

Abstract
The expansion of e-scooter sharing system has led to several novel interactions within the existing transportation system. Although there is a potential for e-scooter sharing and ridesourcing to both compete and complement each other, few studies shed light on the relationship between e-scooter sharing and ridesourcing. To fill this gap, this study uses a LightGBM model and SHAP to explore the spatial heterogeneity of e-scooter’s relationship with ridesourcing. We first collect a list of built environment and social demographic variables related to e-scooter sharing usage. Then we use LightGBM to model the daily e-scooter sharing usage using the daily ridesourcing usage along with the collected built environment and social demographic variables. The model is then interpreted using SHAP to analyze the interactive effects of ridesourcing and spatial variables on e-scooter sharing usage. The results indicate that the contributions of ridesourcing trip count on e-scooter usage are more sensitive in areas with higher population density, fewer young people proportion, and more proportion of people with advanced education level. These findings can assist cities in harmonizing e-scooter sharing and ridesourcing thus promoting sustainable transportation systems.