SWS: an unsupervised trajectory segmentation algorithm based on change detection with interpolation kernels

Trajectory mining aims to provide fundamental insights into decision-making tasks related to moving objects. A fundamental pre-processing step for trajectory mining is trajectory segmentation, where a raw trajectory is divided into several meaningful consecutive sub-sequences. In this work, we propose an unsupervised trajectory segmentation algorithm, Sliding Window Segmentation (SWS), that processes an error signal generated by calculating the deviation of the middle point of an octal window from its imaginary interpolated version. This algorithm is flexible and can be applied to different domains by selecting an appropriate interpolation kernel. We examined our algorithm on three datasets of three different domains such as meteorology, fishing, and people moving in a big city. We also compared SWS with three other trajectory segmentation algorithms, namely GRASP-UTS, CB-SMoT, and SPD. Our experiments show that the proposed algorithm achieves the highest harmonic mean of purity and coverage for all datasets and explored algorithms with statistically significant differences.