Beyond binary relationship: Multivariant analysis between ride-hailing and public transit based on multi-sourcing data

Abstract

The impact of ride-hailing (RH) as an emerging mode of travel service on public transit (PT) systems has been confirmed. However, the current research only views the relationship between PT and RH as competition or complementation based on macro statistics and travel time differences. In fact, the relationship is beyond binary, and it is partial to take the travel time difference as the only classification factor. We constructed a Gaussian mixture model (GMM) using RH data in Xi’an, and three indicators of travel time, cost, and service quality difference were used to classify the relationship between RH and PT. To clarify the factors influencing the relationship classifications, a Multinomial logistic model (MNL) was constructed with the built environment, economic factors, and travel purpose. The results show that the RH-PT relationship can be generally classified into four classifications: Competition (26.5%), RH superiority (47.7%), PT superiority (13.6%), and Irrelevance (12.2%). Competition occurs mainly around metro stations, RH superiority mainly during working hours in outer urban areas, and PT superiority is most widely distributed in the morning peak. POI density and the number of bus lines are positively correlated with Competition, RH superiority, and PT superiority. In addition, there is significant spatial heterogeneity in the RH-PT relationship, for which we constructed a Geographically weighted regression (GWR) model to analyze it. We find that the spatial heterogeneity may stem from the spatial autocorrelation and the spatial disparities in the distribution of regression coefficients. Therefore, policymakers should formulate policies to transform competition from multiple perspectives.