Exploring lifestyle patterns from GPS trajectory data: embedding spatio-temporal context information via geohash and POI

This study explored the growing potential for knowledge discovery from GPS-based trajectory data across various domains, including urban planning and transportation systems. We analyzed the temporal and spatial data from GPS-derived stay points to examine life patterns. Stay points indicate specific geographic locations and provide data on the occurrence and duration of stays. We extracted these stay points from the GPS trajectory data and organized them into temporal sequences. Each sequence was transformed into a vector, incorporating Points of Interest, geohash codes, occurrence times, and duration times. Using representation learning, we reduced the dimensionality of these vectors and applied clustering techniques to identify distinct life pattern. Our analysis revealed five distinct patterns: locally active, student, school-centric, worker, and homebody. This research makes significant contributions to fields such as urban planning by integrating spatial characteristics with temporal and semantic information in life pattern analysis. The application of representation learning has enabled the discovery of meaningful life patterns from GPS-based trajectories.