Periodic Pattern Mining Based on GPS Trajectories

Abstract

With the rise of LBS (Location Based Service), lots of user recommendation system based on location trajectory analysis has emerged. Finding periodic behavior is essential to analysis user's activity. In order to solve the problems with trajectory outliers and human intervention in periodic parameters, we propose a three-stage framework called PPM (Periodic Pattern Mining) to detect periodic pattern based on people's trajectory. First of all, we preprocess the trajectory data to extract stay points. Secondly, the sequence of stay points are clustered to construct the important places. At the last stage, the movement sequence is transformed into a binary sequence. Then period of every binary sequence is detected by a probabilistic model. The experiment based on public mobile dataset shows that the proposed method can be used to mine people's periodic activity pattern effectively.