Community Detection and Location Recommendation Based on LBSN

Community detection is an effective tool for mining hidden information in social networks. Label propagation is a widely used and effective community detection algorithm. A lot of work has been done based on label propagation for standalone machine computing. While in location based social networks (LBSN), paralleled label propagation is needed to deal with large scale datasets. In this paper, we propose paralleled label propagation using MapReduce in Hadoop which improves the efficiency of community detection in the LBSN. We apply community detection to explore the friendship circles among the users and within each friendship circle recommend POI for each user. In the LBSN, we establish the user-location network and define the user-location score to calculate users' similarity. Comparing to direct recommendation, our method improves the accuracy of the recommendation and recall rate, the recommended effect has been improved.