A Novel Alternative Weighted Fuzzy C-Means Algorithm and Cluster Validity Analysis

Abstract

Proposed a novel fuzzy cluster algorithm-AWFCM, aiming at large miss-clustering and invalidation in the fuzzy C-means algorithm when has noises and uneven samples situation. This new algorithm defined a new distance in new metric space and introduced weight matrix based on sample dots' density. New definition of distance can efficiently restrain the error range of clustering centers for samples with noise points in iteration, meanwhile improve recursion for clustering centers according to samples' density. Experiments have proved that AWFCM algorithm overcomes bugs of FCM algorithm to a certain extent, with favorable convergence and robust.