An Empirical Study of Travel Behavior Using Private Car Trajectory Data
Currently, an increasing number of private cars drive on roads and produce a massive volume of trajectory data. These data provide a new opportunity for understanding people's travel behavior. In this paper, we conduct an in-depth empirical study of people's travel behavior using private car trajectory data. First, we conduct the study in terms of four metrics, namely, number of trips, average velocity, trip distance, and entropy, and compare the data with those from taxi and ridesharing. Through experiments, we find that the number of trips, average velocity, and trip distance of private cars are more in line with people's travel habits, and the entropy values of private cars are lower. Therefore, we consider private car trajectory data capable of well characterizing people's travel behavior, even better than that from floating cars (such as taxis and buses) or ridesharing services. Moreover, we reveal that a unique metric (so-called dwell time) from the private car trajectory data helps to further understand people's travel behavior. Finally, we propose a fine-grained destination prediction algorithm as a case study, enhancing the state-of-the-art SubSyn algorithm, and show a good understanding of people's travel behavior facilitate a more personalized prediction of final destinations for private cars.