Real coded genetic algorithm for development of optimal G-K clustering algorithm

Abstract

Clustering has been used as a popular technique for identifying a natural grouping or meaningful partition of a given data set by using a distance or similarity function. This paper proposes a novel real coded Genetic algorithm (GA) for the development of optimal Gustafson Kessel (GK) clustering algorithm. In this work, the objective function of the GK algorithm is optimized using real coded genetic algorithm. The cluster centers are represented as real numbers and real-parameter genetic operators are applied to obtain the optimal cluster centers that minimize the intra-cluster distance. The performance of the proposed approach is demonstrated through three gene expression data sets. Xie-Beni index is used to arrive at the best possible number of clusters. The proposed method has produced the objective function value which is less than the value obtained using K-Means, Fuzzy C-Means and GK algorithms. Statistical analysis of the test results shows the superiority of the proposed algorithm over the existing methods.