Finding frequent sub-trajectories with time constraints

Abstract

With the advent of location-based social media and location-acquisition technologies, trajectory data are becoming more and more ubiquitous in the real world. Trajectory pattern mining has received a lot of attention in recent years. Frequent sub-trajectories, in particular, might contain very usable knowledge. In this paper, we define a new trajectory pattern called frequent sub-trajectories with time constraints (FSTTC) that requires not only the same continuous location sequence but also the similar staying time in each location. We present a two-phase approach to find FSTTCs based on suffix tree. Firstly, we select the spatial information from the trajectories and generate location sequences. Then the suffix tree is adopted to mine out the frequent location sequences. Secondly, we cluster all sub-trajectories with the same frequent location sequence with respect to the staying time using modified DBSCAN algorithm to find the densest clusters. Accordingly, the frequent sub-trajectories with time constraints, represented by the clusters, are identified. Experimental results show that our approach is efficient and can find useful and interesting information from the spatio-temporal trajectories.