A Friend Recommendation Algorithm Based on Multiple Factors in LBSNs

In location-based social networks, the current friend recommendation algorithms just take a relatively single factor into account without comprehensive evaluations. To solve this problem, we design a framework - Multiple Heterogeneous Social Network (MHSN) according to users' profiles, check-in records and interests. Based on this framework, we propose a friend recommendation model which consider multiple factors, including 1) a detecting model based on interest similarity by using users' check-in records, 2) a social distance calculation method based on users' social relationship, 3) a clustering method based on users' check-in location information to measure the similarity among clusters. The top-k friends who satisfy the above conditions will be recommended to the target users. We evaluated our method using Foursquare data-sets and the results showed that our friend recommendation algorithm is more feasible and effective.