Fuzzy clustering with optimized-parameters multiple Gaussian Kernels

Abstract

Kernel fuzzy clustering has been applied to data with nonlinear relationships. Two approaches were used: clustering with a single kernel and clustering with multiple kernels. While clustering with a single kernel doesn't work well with "multiple-density" clusters, Multiple Kernel Fuzzy clustering tries to find an optimal linear weighted combination of kernels with initial fixed (not necessarily the best) parameters. Our algorithm tries to find the optimal parameters of each kernel in each cluster. It is a generalization of the single Kernel Fuzzy c-means and the Multiple Kernel Fuzzy clustering algorithms. In our algorithm, there is no need to give "good" parameters of each kernel and no need to give an initial "good" number of kernels. Every cluster will be characterized by a Gaussian Kernel with optimal parameters. Using different databases and different clustering validity measures, it is compared to other similar clustering algorithms showing its effective clustering performance.