Integration of K-Means Clustering in Genetic Algorithms to Optimize Convergence and Maintain Diversity

Abstract

To mitigate premature convergence in genetic algorithms and to maintain population diversity as iterations increase, a genetic algorithm incorporating K-means clustering is proposed. The integration of the K-means clustering method within the genetic algorithm involves an initial division of the population into distinct subpopulations. Selection, crossover, and mutation operations are then conducted independently within each subpopulation, facilitating local optimization. The optimal value from each subpopulation is preserved, ranked, and the overall optimal value is retained to prevent entrapment in local optima. The effectiveness of this enhanced algorithm is subsequently evaluated using standard test functions.