Federated Rough C-Means Clustering and Its Application to Collaborative Filtering

In clustering-based collaborative filtering (CF), clusters of users with similar preference patterns are extracted, and contents with high preference within the cluster are recommended. In rough clustering-based CF, improved recommendation performance can be expected by considering cluster overlap. However, in the conventional framework, learning is performed by aggregating data from all users on a server, which is undesirable from a privacy protection perspective. In this study, from the viewpoint of privacy protection, we propose federated RCM (F-RCM) clustering by introducing federated learning to rough C-means (RCM) clustering, a type of rough clustering. Furthermore, as an application to CF, we propose F-RCM-CF and verify its recommendation performance through numerical experiments using real-world datasets.