Metadata-Based Collaborative Filtering Using K-Partite Graph for Movie Recommendation

Collaborative filtering recommends items to a user based on the interests of other users having similar preferences. However, high dimensional, sparse data result in poor performance in collaborative filtering. This paper introduces an approach called multiple metadata-based collaborative filtering (MMCF), which utilizes meta-level information to alleviate this problem, e.g., metadata such as genre, director, and actor in the case of movie recommendation. MMCF builds a k-partite graph of users, movies and multiple metadata, and extracts implicit relationships among the metadata and between users and the metadata. Then the implicit relationships are propagated further by applying random walk process in order to alleviate the problem of sparseness in the original data set. The experimental results show substantial improvement over previous approaches on the real Netflix movie dataset.