Exploiting Geographical and Temporal Patterns for Personalized POI Recommendation

Human behavior presents various temporal and geographical patterns that can be used to model user preferences and enhance prediction in the task of POI recommendation. The task of personalized next point-of-interest (POI) recommendation in Location-based Social Networks (LBSNs) has been studied extensively in recent years. The challenge of modeling the interactions of the user, current POI, and next POI presents the need to incorporate sequential dynamics using methodologies like Markov chains and Metric embedding. Existing methods capture these interactions by decomposing them into pairwise relationships. In this thesis, we apply Personalized Ranking Metric Embedding (PRME) for personalized next POI recommendation based on user’s check-in history in various LBSNs like Foursquare and Gowalla. We introduce methods to incorporate spatial and temporal patterns in this metric embedding model. Experiments conducted on the above publicly available datasets indicate superior results demonstrating the effectiveness of incorporating these behavioral patterns in the task of recommendation