Motif discovery based traffic pattern mining in attributed road networks

With the development of intelligent transportation systems, clustering methods are now being adopted for traffic pattern recognition to discover the time-varying laws in road networks; this had attracted significant attention from the industry and academia over the past decades. Existing methods mainly focus on the mobility pattern and spatiotemporal dimension, ignoring the complex relationships among these segments in road networks. The main issues can be divided into two categories: deep integration of the structural and attribute information; global spatial dependencies for clustering structural properties. To address these issues, a clustering method for motif-based attributed road networks is proposed. A higher-order connectivity model based on motif discovery is designed, and a weighted matrix of adjacent segments is defined in the road networks. Moreover, a clustering model for motif-based attributed road networks is constructed, considering the joint relationship between node structure and features. In this study, a set of experiments were conducted on two real-world datasets. The results indicated that the performance of the proposed method is superior to that of the state-of-the-art methods.