Automatic Generation of Fuzzy Membership Functions based on K-Means and EM Clustering

Abstract

Fuzzy partitioning is based on expert opinion in generating fuzzy membership function. Automatic generation of fuzzy partitioning can be made for a given dataset, using different clustering algorithms. We have developed a fuzzy partitioning algorithm, based on clustering algorithm. For medical dataset, automatic generation of fuzzy membership function is done using KMeans and Expected Maximization (EM) clustering algorithms. According to our results, K-Means clustering is more accurate for fuzzy partitioning as compared to EM clustering based on our developed fuzzy partitioning algorithm.