On Relational Fuzzy c-Means and Kernel Fuzzy c-Means Clustering by objectwise β-spread

Abstract

In this report, objectwise β-spread is proposed for the relational fuzzy c-means clustering and the kernelized fuzzy c-means clustering, while in the conventional methods, the revised value β of the given relational data matrix or the given kernel gram matrix is common for the objects in the case that the clustering fails because the relational data matrix is non-Euclidean or the kernel gram matrix is indefinite. By a numerical example, the proposed methods can reduce the summation of the revised values, which leads to keep the feature of the given data more than the conventional methods.