An exploration of improving prediction accuracy by constructing a multi-type clustering based recommendation framework

摘要

Existing clustering-based recommendation methods generally focus on the clustering of users, items or social trust relationships. Although demonstrated to be efficient and scalable to large-scale datasets, these methods are sensitive to the quality of clustering and still suffer from the problem of low accuracy. In order to solve this issue, in this paper, we propose a multi-type clustering based recommendation framework which systematically considers the trust-based user clustering, similarity-based user clustering and similarity-based item clustering to further improve the recommendation accuracy. A SVD (Singular Value Decomposition) signs-based community mining method is utilized to process the trust and distrust matrix in order to discover the trust-based user clusters. The PLSA (Probabilistic Latent Semantic Analysis)-based model is employed to explore the similarity-based user and item clusters. Then a clustering-based trust regularization term is, proposed to incorporate the trust-based user clusters into the matrix factorization model. Comparative experiments on two real-world datasets demonstrate that our approach can better address the issues of data sparsity and cold start, and outperforms other stateof-the-art methods in terms of RMSE and MAE. (C) 2016 Elsevier B.V. All rights reserved.