Predictability analysis on expressway vehicle mobility using electronic toll collection data

Abstract

This paper assesses the predictability of individual vehicle's mobility in Beijing expressway system using Electronic Toll Collection (ETC) data records. By examining the uncertainties of movements using entropy, considering both the frequencies and sequential correlations of vehicles' trajectories, we draw to the conclusion that the average limit of predictability of expressway vehicles mobility is 91%. Furthermore, we concluded that the individual property is negatively correlated to its mobility property such as visited station, average travel distance and radius of gyration. Finally, we applied Markov chain (MC) based models to predict the actual accuracy of predictability and found that MC(1) model is adequate to produce a good result and extending the order of the model won't give substantial bonus, thus the first-order Markov property of expressway vehicles' trajectory is proved. Our findings indicate that individual vehicles' mobility in expressway system is far from random and highly dependent on its historical trajectory