Improving a recommender system by collective matrix factorization with tag information

Collaborative filtering (CF) is the most widely used method of recommender systems. However, it is hard to give users reliable recommendation when there is little information about users. This is the sparsity problem of CF. In this paper, we propose a collective matrix factorization method using tag information to solve the sparsity problem. With tag information, we construct a user-tag matrix that represents users' preferences about tags. Using the user-tag matrix, we convert sparse user-item matrix into dense user-item matrix. In our method, the collective matrix factorization has the role of transferring information between the user-item matrix and user-tag matrix. We experimentally show that our method generates more precise prediction than general CF suffering from the sparsity problem.