Abstract
A fuzzy clustering model for extracting square or cross-shape prototypes in 2-D data spaces was proposed by extending the FCE-type linear clustering method considering local coordinate rotation. In this research, the clustering model is applied to multi-dimensional cases, in which prototype estimation is performed in conjunction with estimation of the 2-D prototypical plane. The new clustering criterion is the linear combination of the distance measure between data points and square/cross-shape prototypes in 2-D planes and the FCV criterion for 2-D prototype estimation.
