Diversified recommendation algorithm for hybrid label based on matrix factorization

As to the problem of Cold-Start and Simplification, a new recommendation algorithm named Diversified Recommendation Algorithm for Hybrid Label based on Matrix Factorization (DRA-HLMF) is proposed in this paper. Firstly, the existing preference of users and relevance of items are excavated in the social networks in order to increase the diversity of label recommendations horizontally and vertically(also users and items); Secondly, the trust value by combining the initial label value with the label popularity can ensure the accuracy of label recommendation. Finally, the recommendation is dynamically updated by the time-weight value. Then the recommendation results are obtained by matrix factorization algorithm. Experiments on real data show that the proposed method can improve the coverage and diversity of label recommendation on the basis of guaranteeing accuracy in comparison with classical algorithms.