Taxi Operation Optimization Based on Big Traffic Data

Motivated by the remarkable improvement of information and communication technologies, along with the rapid progress of urbanization, smart city has become a novel brand. By mining big traffic data generated by widely deployed GPS devices and sensors in modern cities, we can unlock the knowledge of human mobility patterns and social functional regions, and then apply it to tackle critical problems in city construction. One of the tough issues is the paradoxical situation in urban traffic control and management, which is the empty carrying phenomenon for taxi drivers and the difficulty of taking a taxi for passengers. In the paper, we propose a data-driven taxi operation strategy to maximize drivers' profit, reduce energy consumption, and decrease environment pollution. Specifically, we capture social properties of functional areas through integrating, processing and analyzing the big traffic data. Later, we introduce the Time-Location-Sociality model which can identify three dimensional properties of city dynamics to predict the number of passengers in different social functional regions. Furthermore, we recommend Top-N areas for drivers according to the prediction outcomes, which introduce more profitable opportunities to pick up passengers. We conduct extensive experiments using the real GPS data generated by 12,000 taxis during 10 weekdays and 8 weekends in Beijing, and achieve prediction accuracies of 90.14% on weekdays and 86.37% at weekends respectively, which implies the effectiveness of our optimizing taxi operation strategy by considering the three dimensional properties.