Modeling users' vehicles selection behavior in the urban carsharing program

Abstract

Electric carsharing network has been expanding at a very fast rate in the last few years, accompanied by more complex challenges to carsharing operators. Increasing vehicles usage allows for benefits maximization. However, vehicles usage varies significantly in a fleet because of users' preference to vehicles with different features. This study investigated contributing factors to users' vehicles selection behavior through random forests and binary logistic regression using the administrative datasets collected from EVCARD carsharing program. Results showed state of charge (SOC) of electric vehicles and the number of available vehicles parked at a station had the greatest effect on user's vehicles selection behavior. Users tend to be greedy rather than rational when making decisions as they always choose the vehicles with the maximum SOC even their real trips are short. The attributes of trips and users, like real trip distances and users' familiarity with carsharing program, also play an important role in the selection process. Findings from this research can be beneficial to carsharing operators to prioritize investments when purchasing new vehicles and develop optimal management strategies to enhance existing vehicles attraction for users.