Friend Recommendation Algorithm for Online Social Networks Based on Location Preference

In recent years, it has become more and more popular to recommend friends on the location-based social network (LBSN), which is combined with the user's behavior in the real world. LBSN has three attributes including temporal, spatial and social correlation. However, the combination situation of the three cannot be solved in previous algorithms. For instance, the problem of recommending friends with similar location preference in real world cannot be solved by the method based on the social network topology or non-topological information (such as user profile). A new approach that recommends friends with similar location preference for LBSN's users is proposed, in which both the online friendship information and the offline user behavior are considered. The theories and methods including Markov chain, cosine similarity based on location clustering and threshold evaluation are used in the proposed approach. Finally, rationality and effectiveness of the algorithm is verified by using a real dataset which is from a LBSN (Gowalla).