Comparison of User Trajectories Based on Coordinate Data and State Transitions

At present, one can acquire a great deal of trajectory data as positioning technology has prevailed and become sophisticated. In accordance with this fact, data mining technology for extracting important information and knowledge from trajectory data is gaining recognition. Clustering of trajectories and giving meanings to movements are active areas of research. In this paper, we focus on the Markov chain model for representing trajectory data and propose a method for defining states that can cope with the distinguished characteristic of each trajectory and a method for comparing trajectories transitioning over different sets of states in the Markov chain model. From an experiment using real trajectory data, we confirm the advantage of the proposed approach over others.