Personalized POI recommendation based on check-in data and geographical-regional influence

Nowadays, many people like to share the places they visited to their friends in the location-based social networks (LBSNs). Therefore, LBSNs have accumulated large-scale user check-in data and the availability of these data enables many location-based services to users. As a location-based service, point-of-interests (POI) recommendation can provide services to people and promote merchants. Many researchers utilize user-based collaborative filtering and geographical influence for POI recommendation. However, existing studies have two limitations: (1) when modeling user-based CF, users' POI preferences are not fully considered; (2) when modeling geographical influence, geographical features have not been explored deeply. In this paper, we propose a POI recommendation approach by improved user-based CF and geographical-regional influence. Firstly, we construct the user-POI matrix by normalized check-in frequencies, which can effectively represent user preferences. Secondly, we find that each user's check-in POIs can be divided into several regions. Accordingly, we integrate geographical influence with the regional feature to produce recommendation. Finally, we utilize a unified framework to combine improved user-based CF with geographical-regional influence for POI recommendation. Our experimental results on real-world dataset show that the proposed approach outperforms the state-of-the-art POI recommendation methods substantially.