Utilization of Genre Information in Collaborative Filtering by Three-mode Fuzzy Co-Clustering
Katsuhiro Honda, Haruto Miwa, Seiki Ubukata, Akira Notsu · 2024
Recommendation ability of collaborative filtering (CF) is expected to be improved by utilizing not only user-item cooccurrence information but also additional information. In this paper, co-clustering-based CF is improved by introducing three-mode fuzzy co-clustering, where user-item cooccurrence information is utilized in conjunction with additional genre information on each item. User-item co-clusters are extracted such that users' preference tendencies on items are summarized by considering their intrinsic preferences on genre categories, and then, the recommendation capability of co-clustering-based CF can be improved even with sparse cooccurrence information. Experimental results with MovieLens benchmark data demon-strate that recommendation performance is improved by properly increasing the responsibility degree of genre information.