Flow-based unit is better: exploring factors affecting mid-term OD demand of station-based one-way electric carsharing

Abstract

Station-based one-way carsharing (OWC) serves as a flexible method to enjoy the benefits of car travel, while also demonstrating the potential to mitigate environmental challenges and traffic congestion in cities. In the previous studies on OWC, location-based units such as a single station, station cluster, and land parcel were generally used as the basic analysis units for demand; however, these studies failed to consider the association among the operating areas. This led to results that were pertinent to a specific OWC system and cross-section of the development process. The objective of this study is to explore the significant factors related to flow-based demand (i.e., four-weekly bookings from the origin spatial unit to the destination spatial unit (OD bookings)). Four groups of explanatory variables are adopted: carsharing spatial unit attributes, built environment, transportation facilities, and OD trip attributes (such as public transportation and car travel distance between the OD). A combination model integrating machine learning and a generalized linear model is also developed to address the zero-inflation issue of the data. Moreover, an approach of Shapley additive explanations is implemented to determine the considerable effects of the factors. The results show that (1) the OD trip attributes play an important role in estimating the carsharing OD demand; (2) taxi and carsharing demands exhibit a non-significant partial overlap, and carsharing may compete with buses and supplement to the metro; and (3) the travel purpose in carsharing is diverse for any land use over a four-week period.