FCM-type co-clustering with exclusive partition of selected items and application to collaborative filtering

Collaborative Filtering achieves personalized recommendation considering similarities among user preferences and the task has close connections with co-clustering of user-item relational data. FCM-type co-clustering extracts user-item co-clusters, in which users are assigned to clusters in an exclusive manner while item partitions are not necessarily exclusive and each item can be shared (rejected) by multiple (all) clusters. In this paper, an exclusive constraint is introduced into FCM-type co-clustering and the clustering model is demonstrated to be useful in collaborative filtering task.