Online Car-Hailing Trip Purpose Inference Based on Spatiotemporal Attribute

Online car-hailing is gradually becoming an important mode of travel in people's daily life. A lot of location data generated during the ride contains information about the travel, but offers very little about the passenger's trip purpose. However, understanding the trip purpose of passengers has an important significance in practice. On the one hand, it helps to serve passengers better such as providing recommendation services. It also helps the government to plan and provide targeted services better to the public. In this paper, we propose a dual-weighted probability-based model which combines spatial and temporal attributes to infer the trip purpose. We validated this model and its accuracy of inference reached 81.87%. Then, we applied this model on the online car-hailing data of Chengdu area to obtain the trip purposes, and explored the change in the proportion of various trip purposes over time.