Traffic Risk Mining From Heterogeneous Road Statistics

At present, a large amount of traffic-related data is obtained manually and through sensors and social media, e.g., traffic statistics, accident statistics, road information, and users' comments. In this paper, we propose a novel framework for mining traffic risk from such heterogeneous data. Traffic risk refers to the possibility of occurrence of traffic accidents. Specifically, we focus on two issues: 1) predicting the number of accidents on any road or at intersection and 2) clustering roads to identify risk factors for risky road clusters. We present a unified approach for addressing these issues by means of feature-based non-negative matrix factorization (FNMF). In particular, we develop a new multiplicative update algorithm for the FNMF to handle big traffic data. Using real-traffic data in Tokyo, we demonstrate that the proposed algorithm can be used to predict traffic risk at any location more accurately and efficiently than existing methods, and that a number of clusters of risky roads can be identified and characterized by two risk factors. In summary, our work can be regarded as the first step to a new research area of traffic risk mining.