Modeling of Human Movement Behavioral Knowledge from GPS Traces for Categorizing Mobile Users

Human movement analysis and categorization of mobile users based on their movement semantics are challenging tasks. Further, due to security and privacy issues, insufficient labelled or user-annotated data (or, ground-truth data) makes the user-classification from GPS traces more complex. In this work, we present a framework which models user movement patterns containing both spatio-temporal and semantic information, generates semantic stay-point taxonomy by analysing GPS traces of all users, summarizes individuals’ GPS traces and clusters users based on the semantics of their movement patterns. To alleviate labelled data scarcity problem while user categorization in a particular region of interest (ROI), we propose a method to transfer knowledge derived from a set of GPS traces of a geographically distanced but similar type of ROI. An extensive set of experiments using real GPS trace dataset of Kharagpur, India and Dartmouth, Hanover, USA have been carried out to demonstrate the effectiveness of our proposed framework.