CFMT: a collaborative filtering approach based on the nonnegative matrix factorization technique and trust relationships

As a method of information filtering, the Recommender System (RS) has gained considerable popularity because of its efficiency and provision of the most superior numbers of useful items. A recommender system is a proposed solution to the information overload problem in social media and algorithms. Collaborative Filtering (CF) is a practical approach to the recommendation; however, it is characterized by cold start and data sparsity, the most severe barriers against providing accurate recommendations. Rating matrices are finely represented by Nonnegative Matrix Factorization (NMF) models, fundamental models in CF-based RSs. However, most NMF methods do not provide reasonable accuracy due to the dispersion of the rating matrix. As a result of the sparsity of data and problems concerning the cold start, information on the trust network among users is further utilized to elevate RS performance. Therefore, this study suggests a novel trust-based matrix factorization technique referred to as CFMT, which uses the social network data in the recommendation process by modeling user's roles as trustees and trusters, given the trust network's structural information. The proposed method seeks to lower the sparsity of the data and the cold start problem by integrating information sources including ratings and trust statements into the recommendation model, an attempt by which significant superiority over state-of-the-art approaches is demonstrated an empirical examination of real-world datasets.