The role of built environment in shaping reserved ride-hailing services: Insights from interpretable machine learning approach

Abstract

Cruising ride-hailing vehicles exacerbate traffic congestion by generating negative externalities. In contrast, reserved ride-hailing services leverage precise information regarding the departure times and origins-destinations of future trips. Platforms can use this data to dispatch and route drivers more efficiently, thereby reducing the need for cruising. Although previous research has largely concentrated on real-time ride-hailing services, the impact of the built environment on reserved ride-hailing remains unexplored with empirical data. This study integrates multi-source data from Haikou City in China and utilizes the gradient boosting decision tree model, which is an interpretable machine learning approach, to investigate potential relationships between reserved ride-hailing trip demand and the built environment. The rankings of relative importance reveal that factors such as the density of food services, education institutions, accessibility to town centers, and proximity to transportation hubs significantly influence the demand for reserved ride-hailing. Furthermore, the study demonstrates that the aforementioned factors exhibit non-linear effects on the demand for reserved ride-hailing. The findings have policy implications for local governments aiming to promote reserved ride-hailing and enhance urban mobility services.