Clustering of Mountain Hiking GPS-Trajectory Data

The use of GPS applications in exercise and leisure activities has been increasing. While it is believed that exercise patterns of GPS application users are included in spatio-temporal GPS trajectory data, analyzing these data is not easy due to its large volume and unstructured nature. In this paper, we suggest a data-adaptive clustering method to analyze the characteristics of mountain tracking patterns. To reflect complex spatio-temporal patterns of GPS-trajectories, the proposed clustering method based on a new similarity measure, which is an weighted average of space, time, and velocity similarities. Furthermore, we suggest a an adaptive similarity weight selection strategy by considering a grid search with appropriate constraints. When we applied the clustering analysis methodology we proposed to the Mount Worak GPS-sampling data, results show that spatio-temporal trajectory information has a close relationship with individual characteristics which are unused in the clustering analysis. These findings suggest that the possibility of providing a system that allows GPS application providers and users to recommend hiking trails or check personalized hiking preferences would be beneficial.