CBGDC: A new genetic center based data clustering algorithm based on K-means

In this paper, a Center Based Genetic Data Clustering (CBGDC) algorithm based on K-means is proposed. This algorithm is able to detect arbitrary shape clusters and will not converge to local optima. In proposed algorithm a new population initialization method and reinsertion way have been used. Crossover and mutation operators will not be done with a fix probability and a new fitness function based on Silhouette index will be used toevaluate fitness of chromosomes faster. The efficiency of CBGDC has been compared with original genetic data clustering and K-means algorithm on artificial and real life datasets and experimental results show that the CBGDC will decrease clustering error more than original genetic data clustering and K-means.

Keywords:

Data mining , data clustering , genetic algorithm , partitioning , K-means algorithm