Evaluation of Social, Geography, Location Effects for Point-of-Interest Recommendation

Abstract

Recently, location-based social network services have become very popular. Therefore, point of interests (POIs) recommendation has also become a promising and hot research problem. In POIs recommendation, the number of locations could be more than the number of users, so it is a challenge to recommend relevant locations. In this paper, we incorporate user preference, social influence and attraction of locations in the recommendation. First, we use geographic influence for candidate selection. Furthermore, we propose a unified POI recommendation framework, which fuses user preference to a POIs with social influence and attraction of locations methods based on customized linear weighting. In addition, we discuss performance of classification-based models (logistic regression and libFM) for POI recommendation. Experimental results show the unified POI recommendation framework based on customized linear weighting outperforms other approaches.