Road traffic prediction by incorporating online information

Tian Zhou, Lixin Gao, Daiheng Ni

Abstract: Road traffic conditions are typically affected by events such as extreme weather or sport games. With the advance of Web, events and weather conditions can be readily retrieved in real-time. In this paper, we propose a traffic condition prediction system incorporating both online and offline information. RFID-based system has been deployed for monitoring road traffic. By incorporating data from both road traffic monitoring system and online information, we propose a hierarchical Bayesian network to predict road traffic condition. Using historical data, we establish a hierarchical Bayesian network to characterize the relationships among events and road traffic conditions. To evaluate the model, we use the traffic data collected in Western Massachusetts as well as online information about events and weather. Our proposed prediction achieves an accuracy of 93% overall.