GPS Trajectory segmentation and clustering method

Abstract

The analysis of GPS trajectories, in the context of current life, constitutes one of the main sources of support for decision making. The detection of patterns in GPS trajectories that allow determining actions to follow and decisions to be taken more quickly is one of the main lines of research in this field. In this paper we propose a method of segmentation and clustering of GPS trajectories composed of a segmentation algorithm that aims to segment the trajectories based on a partitioning criterion. Then the generated segments are used by a trajectory clustering algorithm based on traditional k-means; the method improves the clustering centroid selection process by using an initial grid and proposes a similarity function between trajectory segments for the clustering process. The results of the experiments performed show a significant improvement compared to the methods published in the literature.