Location-Based Social Networks: Latent Topics Mining and Hybrid Trust-Based Recommendation\ud

The rapid advances of the 4th generation mobile networks, social media and the ubiquity of the advanced mobile devices in which GPS modules are embedded have enabled the location-based services, especially the Location-Based Social Networks (LBSNs) such as Foursquare and Facebook Places. LBSNs have been attracting more and more users by providing services that integrate social activities with geographic information. In LBSNs, a user can explore places of interests around his current location, check in at these venues and also selectively share his check-ins with the public or his friends. LBSNs have accumulated large amounts of information related to personal or social activities along with their associated location information. Analyzing and mining LBSN information are important to understand human preferences related to locations and their mobility patterns. Therefore, in this thesis, we aim to understand the human mobility behavior and patterns based on huge amounts of information available on LBSNs and provide a hybrid trust-based POI recommendation for LBSN users.\ud In this dissertation, we first carry out a comprehensive and quantitative analysis about venue popularity based on a cumulative dataset collected from greater Pittsburgh area in Foursquare. It provides a general understanding of the online population's preferences on locations. Then, we employ a probabilistic graphical model to mine the check-in dataset to discover the local geographic topics that capture the potential and intrinsic relations among the locations in accordance with users' check-in histories. We also investigate the local geographic topics with different temporal aspects. Moreover, we explore the geographic topics based on travelers' check-ins. The proposed approach for mining the latent geographic topics successfully addresses the challenges of understanding location preferences of groups of users. Lastly, we focus on individual user's preferences of locations and propose a hybrid trust-based POI recommendation algorithm in this thesis. The proposed approach integrates the trust based on both users' social relationship and users' check-in behavior to provide POI recommendations. We implement the proposed hybrid trust-based recommendation algorithm and evaluate it based on the Foursquare dataset and the experimental results show good performances of our proposed algorithm.\u