Matrix Factorization Meets Cosine Similarity: Addressing Sparsity Problem in Collaborative Filtering Recommender System

摘要: Matrix factorization (MF) technique has been widely used in collaborative filtering recommendation systems. However, MF still suffers from data sparsity problem. Although previous studies bring auxiliary to solve this problem, is not always available. In paper, we propose a novel method, Cosine Factorization (CosMF), address the problem without data. We observe that when sparse, magnitude of user/item vector could be properly learned due lack information. Based on observation, use cosine replace inner product for sparse users/items, thus eliminating negative effects poorly trained magnitudes. Experiments various real life datasets demonstrate CosMF yields significantly better results help dataset.