Modified clustering-based differential evolution with a flexible combination of exploration and exploitation

Differential evolution (DE) has been extensively used in optimization problem. However, original DE has some shortcomings. Up to now, there have been a lot of its variations. In this paper, a modified version of differential evolution algorithm is raised on the basis of clustering-based differential evolution with random-based sampling and Gaussian sampling. The modified one is called MGRCDE. It can enhance the ability of searching for final solution with better quality by maintaining the diversity of population and local search around individuals with the best quality in the subpopulation. At the same time, it accelerates convergence rate of evolution process by clustering. Twenty-five standard, unconstrained single-objective benchmark functions have been used in verifying the performance of the modified algorithm, and a comparison between the modified algorithm and the previous one has been made. The results demonstrate that the modified algorithm can control the population to move toward global optimal point more effectively, having a better ability of global optimization. Especially in high-dimensional functions, the advantage has been proved more obvious.