Identifying individuals with similar mobility patterns via the conformal prediction framework

• Main Author: Malhotra, Nishtha
• Other Authors: Ho Shen-Shyang
• Format: Final Year Project
• Language: English
• Published: 2015
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: http://hdl.handle.net/10356/62811
• Institution: Nanyang Technological University

In this paper we investigate the GPS trajectories of the users for similarity exploration. GPS trajectories are in the form of time-dependent or independent regions or areas or sequence of regions represented by a sequence of time-stamped points, each of which contains the information of latitude, longitude and altitude. These spatiotemporal patterns are the foundation of many important and practical applications such as traffic analysis, behavior analysis, etc. We extract the spatiotemporal patterns and investigate how the conformal prediction framework can be used for similarity exploration of users and personalized friend and location recommendation. Spatiotemporal patterns containing GPS points are converted to Points of Interest trajectory. The novelty in this approach is that we can compare the geographic and semantic patterns of the users based in different locations using the POI trajectories. Hence two users visiting similar points of interest in different geographic locations can be considered to have similar liking and disliking. A hierarchical graph framework is constructed to uniformly model each individual’s location history and efficiently measure the similarity among users. The deeper levels of the graph model the daily behavior of the users. The distance metric used for comparing trajectories is based on the Longest Common Subsequence algorithm. The reliability of the predictions is ensured by using conformal prediction. After analyzing the user’s behavior and the preferences based on their location history a personalized friends and location recommender system is developed. The trajectories of the friends of the user are analyzed to recommend the locations to the user.