Clustering by genetic algorithm: High quality chromosome selection for initial population

Many existing GA based clustering techniques generate the number of genes for a chromosome through random selection. Due the random selection there is a high chance of getting poor quality of initial genes in the initial population. A poor-quality initial population is likely to produce a poor-quality clustering solution. We argue that having a set of high quality chromosomes in the initial population we are more likely to produce a clustering solution of higher quality. Therefore, in this paper we propose a genetic algorithm based clustering technique that produces high quality initial chromosomes. The proposed technique selects the first 50% of the chromosomes through a deterministic selection phase and the remaining 50% chromosomes through a random selection phase, for the initial population. The proposed technique also uses crossover and mutation operation to getting better clustering result. We conduct experiments on seven datasets that are available in UCI machine learning repository. Two evaluation criteria namely silhouette coefficient and DB index are used. Our experiment results, based on the two evaluation criteria indicate a clear superiority of our technique over three existing techniques namely AGCUK, GAGR and K-means.