Group-Based Personalized Location Recommendation on Social Networks

Location-based social networks (LBSNs) have attracted significant attention recently, thanks to modern smartphones and Mobile Internet, which make it convenient to capture a user’s location and share users’ locations. LBSNs generate large amount of user generated content (UGC), including both location histories and social relationships, and provide us with opportunities to enable location-aware recommendation. Existing methods focus either on recommendation efficiency at the expense of low quality or on recommendation quality at the cost of low efficiency. To address these limitations, in this paper we propose a group-based personalized location recommendation system, which can provide users with most interested locations, based on their personal preferences and social connections. We adopt a two-step method to make a trade-off between recommendation efficiency and quality. We first construct a hierarchy for locations based on their categories and group users based on their locations and the hierarchy. Then for each user, we identify her most relevant group and use the users in the group to recommend interested locations for the user. We have implemented our method and compared with existing approaches. Experimental results on real-world datasets show that our method achieves good quality and high performance and outperforms existing approaches.