Successive POI Recommendation with Category Transition and Temporal Influence

With the popularization of mobile devices and wireless networks, people are able to share their experience on points of interest (POIs) in social networks through "check-ins." Therefore, the problem of successive POI recommendation has been proposed to recommend some POIs to users so that the users are likely to check in at these POIs in the near future. In this paper, we propose a two-phase method to solve the problem of successive POI recommendation. First, we utilize the Matrix Factorization technique to analyze the interaction of users and their sequential check-in behavior with time influence and POI categories, and select the candidate categories that the user will visit. Then, after removing those POIs not belonging to the candidate categories, we fuse user preferences, temporal influence and geographical influence together and finally recommend the POIs with high scores to users. The experimental results on a real check-in dataset show that our recommendation method is better than several state-of-the-art methods in terms of precision and recall.