Investigating the Determinants of Competition Between Ride-Hailing Services and Public Transit

Abstract: Ride-hailing has transformed the landscape of passenger transportation systems in cities worldwide; however, the competition between ride-hailing and public transportation remains a topic of active debate. Additionally, the nonlinear and spatial effects of the determinants influencing this competition have not been fully captured. With the advancements in eXplainable Artificial Intelligence (XAI) techniques, such as SHapley Additive exPlanations (SHAP), machine learning (ML) and artificial intelligence (AI) are becoming increasingly interpretable. This progress offers a unique opportunity to effectively interpret the complex nonlinear effects of determinants and understand their spatial variations. In this study, we apply SHAP in combination with the ML algorithm eXtreme Gradient Boosting (XGBoost) to analyze the nonlinear and spatial effects—including spatial autocorrelation and spatial heterogeneity—of the determinants on the competition between ride-hailing services and public transportation. The results demonstrate that the interpretable ML approach outperforms traditional spatial econometric models, such as the Spatial Error Model (SEM), Spatial Lag Model (SLM), Geographically Weighted Regression (GWR), and Multiscale Geographically Weighted Regression (MGWR). The findings provide valuable insights to foster greater integration between shared mobility and public transportation. In addition, this research highlights the increasing relevance of XAI techniques in transportation studies, showcasing their potential to enhance the interpretability of complex ML models and support evidence-based decision-making in multimodal transportation planning.