Exploring the built environment impacts on Online Car-hailing waiting time: An empirical study in Beijing

Waiting time is a critical metric for evaluating the quality of online car-hailing services, with studies indicating a close association between Online Car-hailing waiting time (CWT) and the urban built environment (BE). Points of interest (POI) data is widely utilized to characterize the built environment. However, quantifying the proper combination or spatial relationships between different types of POIs to capture the functional features of the built environment within a region poses a challenge. This paper proposes a framework for analyzing built environment characteristics based on a semantic probabilistic topic model utilizing Latent Dirichlet Allocation, demonstrating that integrating the building area of POIs along with the travel activity intensity at their respective locations enables a more precise identification of regional functions. Moreover, while some studies have explored the correlation between the built environment and waiting time, few have evaluated the nonlinear interactions between them. On this foundation, employing the machine learning technique XGBoost model in conjunction with online car-hailing order data, we probe the relationship between the built environment and waiting time. The research indicates that CWT is comprehensively affected by multiple factors. Taking weekday evening peak period as an example, the CWT of Central-urban area, where has dense commercial and office land use, positively correlated with commercial and office topics, while negatively correlated with educational, residential, leisure and healthcare topics, leading to a longer CWT in the Central-urban. In addition, the interaction between BEs can weakens their individual effects on CWT. A higher degree of job-residence balance can mitigate the negative impact of residential and office topics on increasing CWT, particularly in Intermediate-urban areas. Additionally, BE topics may also suppress the positive effect of road density on reducing CWT. The relationship between the BE and CWT exhibits threshold effects and a V-shaped relationship, indicating that the BE is significantly associated with CWT only within specific ranges. This correlation also exhibits a gradient pattern from the Central-urban to the Sub-urban, especially concerning office and residential topics. These outcomes elucidate the salient ranges of the built environment that exert substantial impacts on waiting time, informing strategic planning for online car-hailing dispatch and urban development to augment passenger travel satisfaction.