Project Overview

Project information

  • Category: Framework Design/ System
  • Project date: 2019-2020

With the huge volume of collected GPS data, another interesting and obvious question arises that whether all the GPS points have same information or some of the GPS points carry more contextual information than rest. To this end, we aim to model and efficiently store user-movement traces without losing any significant information. However, instead of raw GPS log (time-stamped latitude, longitude data) human movement patterns are better understood when some contextual information, landmarks on the route, duration of stay points or activities performed at stay points are considered. Motivated by the potential merits, recently research trend is to extract semantic information or capturing inherent meaning of these huge volume of human movement data. This semantic enrichment of raw GPS log bridges the gap between collected GPS traces and various location based applications. We aim to capture behavioral differences in the movement patterns of the individuals and utilize the knowledge to cluster users having similar movement patterns. The problem becomes quite challenging for the unpredictable behavior of human mobility. For example, user X and user Y both are students, but may take different routes to university. To tackle this challenge, we propose Bayesian network for modelling user’s movement pattern which captures probabilistic measures of the randomness of movement. Although advances in location-acquisition techniques have generated huge amount of GPS data, but unfortunately, scarcity of user-annotated or labelled data is still a major challenge in categorization of users. Therefore, we aim to learn movement patterns of a known region (source) and map the knowledge of the source region to another region (destination) of same domain (say, academic, commercial etc.). This problem of trajectory knowledge transfer has not been reported well in the existing literature. To address the above mentioned challenges and issues, we propose a framework which involves (i) generating user trajectory from raw GPS log (ii) creating User Trace Model to represent individuals movement behavior, generating place knowledge base of the region from the GPS traces (iii) spatiotemporal movement pattern mining and similarity measurement (iv) transfer the human movement behavior knowledge to other geographical place.