Using User-Item Sub-Block to Improve Recommendation Systems

As an indispensable technique in the field of information filtering, recommendation systems (RSs) have been well studied and developed both in academia and in industry recently. In this paper, we propose the intimacy among users to obtain a user-item objective rating matrix, which can reflect user’s real interest. For the sake of better predicting ratings, a user-item sub-block is presented to cluster a group of intimate users and a subset of items. Then, the sub-block can be detected through intimacy among users and similarity between items. In order to improve recommendation accuracy, we propose a social contribution degree and social similarity based matrix factorization method to predict scores in sub-block. The final predicted ratings are obtained by combining all sub-blocks. Top- N items with highest predicted scores are recommended to each user. Systematic simulations on real world data set have demonstrated the effectiveness of our proposed approach.