A clustering-based differential evolution with random-based sampling and Gaussian sampling

Differential Evolution (DE) has been widely researched because of its excellent performance and many differential evolution variants have been proposed. However, no variant was able to consistently perform over a wide range of test problems. This paper presents a novel algorithm based on the one-step k-means clustering, random-based sampling and Gaussian sampling to improve the performance of DE to solve optimization problems efficiently. The proposed enhanced DE utilizes the one-step k-means clustering to generate k search spaces. In these spaces, the new mutation operators based on random-based sampling and Gaussian sampling are used to exploit. The resulting algorithms are named as clustering-based differential evolution with random-based sampling and Gaussian sampling (GRCDE). Experimental verifications are conducted on 25 benchmark functions and the CEC׳05 competition, including detailed analysis for GRCDE. The results clearly show that GRCDE outperforms other state-of-the-art evolutionary algorithms in terms of the solution accuracy and the convergence rate.