FCM-Type Fuzzy Clustering of Mixed Databases Considering Nominal Variable Quantification
Katsuhiro Honda, Ryo Uesugi, Hidetomo Ichihashi · Journal of Advanced Computational Intelligence and Intelligent Informatics · 2007
This paper proposes a clustering algorithm that performs FCM-type clustering of datasets including categorical data. The proposed algorithm iterates categorical data quantification in FCE clustering so that quantified scores suit the current fuzzy partition. The objective function is the linear combination of two cost functions, i.e., the objective function of FCE clustering and the clustering criterion of quantified category scores. Because quantified category scores are assigned considering the relationship among categories, they are useful for interpreting the cluster structure.