Leveraging tagging and rating for recommendation: RMF meets weighted diffusion on tripartite graphs

摘要: Abstract Recommender systems (RSs) have been a widely exploited approach to solving the information overload problem. However, performance is still limited due extreme sparsity of rating data. With popularity Web 2.0, social tagging system provides more external improve recommendation accuracy. Although some existing approaches combine matrix factorization models with tag co-occurrence and context tags, they neglect issue that would also result in inaccurate recommendations. Consequently, this paper, we propose novel hybrid collaborative filtering model named WUDiff_RMF, which improves regularized (RMF) by integrating Weighted User-Diffusion-based CF algorithm(WUDiff) obtains similar users from weighted tripartite user–item–tag graph. This aims capture degree correlation network enhance recommendation. Experiments conducted on four real-world datasets demonstrate our significantly performs better than already used methods accuracy Moreover, results show WUDiff_RMF can alleviate data sparsity, especially circumstance made few ratings tags.