Robust fuzzy Co-clustering algorithm
William-Chandra Tjhi, Lihui Chen · 2007
Co-clustering is a simultaneous clustering of objects and its features, and is known to be effective for categorization of high-dimensional data. Fuzzy co-clustering is co-clustering in which the resulting co-clusters are represented by fuzzy sets. We introduce a new robust fuzzy co-clustering algorithm called Robust Fuzzy Co-clustering (RFCC). Existing prominent fuzzy co-clustering algorithms rely solely on an Fuzzy C-means-like fuzzy object membership, which is known to be vulnerable to outliers. In RFCC, we propose to incorporate an additional and more robust type of fuzzy object membership to reduce the sensitivity of fuzzy co-clustering to outliers. In this paper, we detail the formulation of RFCC and demonstrate its effectiveness through an experiment on an artificial dataset.