An Annotating Method of GPS Trajectory Data for Human Mobility Analysis in Urban Area

The widespread use of GPS-enabled devices enables us to accumulate unprecedented amounts of individuals' trajec-tories with high accuracy, providing an abundant data resource for human mobility research. We propose a grid-based distance measure between trajectories and an annotation-aware clustering method. We annotate the trajectory by considering the moving mode of humans (walk, trip, or stay) inferred from the moving speed calculated from the trajectory data. We model a one-day human trajectory as a path that consists of a sequence of multiple segments connecting Point-of-Interests (PoIs) associated with episodes. We apply our method to a real-world dataset of human trajectories in the Tokyo area. We confirmed that discriminating the moving mode (walk and trip) of segments yielded meaningful clusters. We found that the smaller grid size of 50 m square produced finer resolution clusters than the larger grid size of 100 m square.