Sequential Extraction by Using Two Types of Crisp Possibilistic Clustering

Possibilistic clustering is well-known as one of the useful clustering methods because it is robust against noise or outlier in data. In the previous study, sparse possibilistic clustering and its variant has been proposed by using L1-regularization. These possibilistic clustering methods with L1-regularization are quite different from the viewpoint of membership function. Two types of new possibilistic approach with L1-regularization named crisp possibilistic clustering are proposed in this paper. Classification function of proposed methods which shows allocation rule in whole space and the way of sequential cluster extraction are also proposed. The effectiveness of proposed methods is, moreover, shown through numerical examples.