A Collaborative Filtering Model based on Matrix Factorization and Trust Information

Abstract

Collaborative filtering (CF) approach is the most efficiency technique for employing as the recommendation system engine. One of the notable types of CF techniques are Singular Value Decomposition (SVD) based techniques. Off-line learning recommendation models use the historical user rating to extract the knowledge. Nevertheless, the difficulties of CF approach are caused from the data sparsity and cold start problems. Aiming to deal with these problems, this research proposed a novel collaborative filtering model relied on matrix factorization technique by incorporate user's trust information into the explicit historical rating scores of users on the items for generating the rating scores prediction model, called Trust-rRSVD. The empirical experiment was established. The accuracy results shown that Trust-rRVSD performed better than the other techniques along with better mitigating the above problems.