IMPROVING FUZZY C-MEANS THROUGH BIASING FUZZY MEMBERSHIP MATRIX

Abstract

Clustering is one of the important data mining tasks. It is the process of partitioning dataset into clusters so that similar objects are placed into the same cluster while dissimilar objects are placed into different clusters. Many clustering techniques have been developed and studied. Based on different theories and methodologies, their types include partitional, hierarchical, density-based, grid-based and model-based clustering algorithms. A partitional clustering algorithm Fuzzy c-means is a well-known and widely used algorithm for data clustering. Fuzzy c-means uses a fuzzy membership matrix U to represent the degree of membership for data points with the clusters. In this thesis, a new step is introduced to Fuzzy c-means clustering algorithm. In this new step, the fuzzy membership matrix is biased through a suggested multiplying factor to improve its accuracy. Experimental results on both synthetic and real data have shown the better performance attained by improved Fuzzy c-means in comparison to classical Fuzzy c-means algorithm.