A Review Paper of Model Based Collaborative Filtering Techniques

The development of Web 2.0 has seen a growth in rapid data and has caused an information overload problem. Recently, collaborative filtering of Recommendation System (CFRS) has attracted many researchers to handle this problem. Several survey papers have been proposed on CFRS, which concentrate on specific aspects of RSs, such as explanations of various challenges, methods or supported sources of RSs, none of them includes all of these aspects in one paper. Thus, this study aims to overcome the aforementioned limitation and collect the most significant aspects of CFRS in one paper to facilitate the researchers' tasks. First, the most dominant approaches involving the data mining methods are presented. Second, seven different challenges that are affect the prediction performance negatively are illustrated, including data sparsity, cold-start, scalability, synonymy, privacy-preserving, gray sheep users, and shilling attacks. Eventually, recent research directions such as social-based, session-based, multi-criteria and mobile-based were highlighted.