A Clustering-Based Matrix Factorization Method To Improve The Accuracy Of Recommendation Systems

Matrix approximation is a common model-based approach to collaborative filtering in recommender systems. However, due to data sparsity, it is difficult for current approaches to accurately approximate unknown rating values, which may cause low-quality recommendations. In this paper, we proposed a modified latent factor model to predict the missing ratings and generate accurate recommendations. The proposed method is able to overcome data sparsity and also improving matrix approximation by integrating clustering and transfer learning techniques in a unified framework. The performance of the proposed method was evaluated on two real-world benchmarks and results show its superiority compare to the state-of-the-art methods.