Friend Recommendation by User Similarity Graph Based on Interest in Social Tagging Systems

Social tagging system has become a hot research topic due to the prevalence of Web2.0 during the past few years. These systems can provide users effective ways to collaboratively annotate and organize items with their own tags. However, the flexibility of annotation brings with large numbers of redundant tags. It is a very difficult task to find users' interest exactly and recommend proper friends to users in social tagging systems. In this paper, we propose a Friend Recommendation algorithm by User similarity Graph (FRUG) to find potential friends with the same interest in social tagging systems. To alleviate the problem of tag redundancy, we utilize Latent Dirichlet Allocation (LDA) to obtain users' interest topics. Moreover, we propose a novel multiview users' similarity measure method to calculate similarity from users' interest topics, co-collected items and co-annotated tags. Then, based on the users' similarities, we build user similarity graph and make interest-based user recommendation by mining the graph. The experimental results on tagging dataset of Delicious validate the good performance of FRUG in terms of precision and recall.