A Novel Approach to Trajectory Analysis Using String Matching and Clustering

Clustering of sub-trajectories is a very useful method to extract important information from vast amounts of trajectory data. Existing trajectory clustering algorithms have focused on geometric properties and spatial features of trajectories and sub-trajectories. In contrast to the existing trajectory clustering algorithms, we propose a new framework to cluster sub-trajectories based on a combination of their spatial and non-spatial features. This algorithm combines techniques from grid based approaches, spatial geometry and string processing. First, we convert each trajectory into a representative sequence that captures the trajectory direction and location. We identify common sub-trajectories from all the sequences using a modified string matching algorithm. Then, we extract non-spatial features from the common sub-trajectories. Finally, we present a density based clustering algorithm to cluster the sub-trajectories. Experimental results show that our framework correctly discovers groups of similar sub-trajectories with their similar non-spatial features.