Square-shape Cluster Extraction by FCE with Local Coordinate Rotation

Abstract

This paper proposes an approach for extracting square-shape clusters, in which the FCE algorithm is applied in conjunction with local coordinate rotation. By rotating data points around the centers of squares, four sides of prototypical squares are estimated as a single linear sub-structure. After the rotation phase, each linear sub-structure is extracted in a similar procedure with local minor component analysis. Because FCE also considers optimization of the FCM objective function, the proposed algorithm can estimate not only square-shape clusters but also grid-like clusters having long tails by changing the priority weight of FCM.