Discovering Traffic Outlier Causal Relationship Based on Anomalous DAG

摘要

The increasing availability of large-scale trajectory data provides us more opportunities for traffic pattern analysis. Nowadays, outlier causal relationship among traffic anomalies has attracted a lot of attention in the research of traffic anomaly detection. In this paper, we propose a model of constructing anomalous directed acyclic graph (DAG) which is based on spatial-temporal density to detect outlier causal relationship in traffic. To the best of our knowledge, the graph theory of DAG is firstly used in this area and the algorithm with strong pruning is proved to have lower time complexity. Moreover, the multi-causes analysis helps reflect the causal relationship more precisely. The advantages and strengths are validated by experiments using large-scale taxi GPS data in the urban area.