A Potential Friend Recommendation Algorithm for Obtaining Spatial Information

With the rapid development of social network, friend recommendation algorithm has become an important component of social application. Location-based social network (LBSN) enables users to record and share their locations anytime and anywhere, which is a high quality information source. In order to meet people's demand of expanding social circle and obtaining diversified spatial information when making friends, this paper proposes a potential friend recommendation algorithm based on the similarity of user's check-in behavior and spatial information acquisition level in the real world. Firstly, we employ kernel density estimation and time entropy to solve the problems of data sparsity and low concentration, then employ cosine distance to measure the check-in behavior similarity. Secondly, we analyze users’ spatial distribution of check-in location and cognitive differences on spatial information. Finally, the method mentioned above is tested with dataset called Foursquare. The results of the experiment show that the proposed method has competitive performance.