Item membership fuzzification in fuzzy co-clustering based on multinomial mixture concept
Katsuhiro Honda, Shunnya Oshio, Akira Notsu · 2014
Co-clustering is a promising technique for summarizing cooccurrence information such as purchase history transactions and document-keyword frequencies. A close connection between fuzzy c-means (FCM) and Gaussian mixture models (GMMs) have been discussed and several extended FCM algorithms, which are induced by the GMMs concept, were proposed. Multinomial mixture models (MMMs) is a probabilistic model for co-clustering task and we have a possibility of inducing a fuzzy co-clustering model based on the MMMs concept, whose goal is to simultaneously estimate the cluster membership degrees of both objects and items. In this paper, a fuzzification mechanism for item memberships is proposed and its characteristic features are discussed.