Enhancing Multi-factor Friend Recommendation in Location-based Social Networks

Recently, location features have become available by the most popular online social networks such as Facebook, Twitter, and Foursquare. These networks are called location-based social networks (LBSN), which allow users to share their locations and location-related content. One of the services that LBSNs present is friend recommendation. This service recommends new friends to users based on their posts, media, opinions, locations or social ties. Several studies have been conducted in the area of friend recommendation by LBSNs. Nevertheless, addressing users' opinions on different topics, and considering asymmetric (directed) relations as one indirect relation, are factors that have not been extensively reviewed. In this paper, we propose a new recommendation model in LBSNs for recommending friends to users based on social, spatial, and textual features. These features are extracted based on factors to enhance the accuracy of friend recommendation process. We propose a novel factor for the extraction of the textual feature, also, we enhance factors for social and spatial features. The recommendation model was tested and evaluated using two real datasets from Twitter. The results of recommending top-k friends showed an average accuracy of 97.88%, and the experiments also showed the efficiency of the recommendation model against another friend recommendation model.