Initial Centroid Determination Using Genetic Algorithm in Data Clustering

Abstract

Clustering K-Means using random initial determination centroid. Generated random centroids using K-Means trapped in optimum local which results in poor clustering quality. Initial centroids in k-means will examine effect of genetic algorithms are each tested on data with dimension reduction and without dimension reduction. Based on the results of initial centroid testing obtained from genetic algorithms, quality of cluster results increase 54.9% in high dimensional data and 52.4% in data had been carried out for dimensional reduction. This shows that K-Means clustering with initial centroids obtained from genetic algorithm calculations has best cluster with significant results.