Abstract

With the rapid adoption of wireless sensor networks (WSNs) into smart cities and vehicle networks, traffic problems can be evaluated and predicted in real-time. In this paper, we propose a data-driven approach to find out the most influential causes of traffic congestions. We first find the top most influential regions and use the Fortune’s algorithm to partition the city. Second, we propose a model with three correlations to measure the dependency between two traffic events, which are spatial correlation, temporal correlation, and logical correlation. Third, we adapt the Independent Cascade model with a pruning algorithm to address traffic congestions. At last, we conduct intensive experiments on large real-world GPS trajectories generated by more than 10,200 taxis in Shanghai to demonstrate the performance of our approaches.