Fuzzy-Possibilistic Clustering for Categorical Multivariate Data

Fuzzy co-clustering is a technique for extracting co-cluster structures from cooccurrence data among objects and items, where fuzzy partition of objects is associated with possibilistic partition of items. In this paper, with the goal of improving the noise robustness of object fuzzy partition, fuzzy-possibilistic partition is introduced into fuzzy clustering for categorical multivariate data (FCCM). Not only the conventional probabilistic fuzzy memberships but also possibilistic typicality memberships are jointly utilized in estimation of object partition while item partition is still remained to be possibilistic. The characteristics of the proposed algorithm are demonstrated through numerical experiments.