One of the challenges in science, especially in computational science, is the optimization problem. Therefore, different methods have been presented for assessment and resolving of these problems. One of the methods which has recently been in the center of attention is evolutionary optimization method. This method attempts to resolve optimization problems by mimicking the mechanisms of biological evolution and behavior of organisms. Evolutionary optimization algorithms are usually subject to defects and shortcomings which make the evolution processes complicated or even prevent them from reaching the global optimum. One of the aforementioned defects is the high number of function evaluations leading to the desired solution. This defect becomes especially obvious where the optimization problem is complex. However, recent research on using machine learning methods, in particular surrogate models in evolutionary optimization show that this combination (machine learning with evolutionary algorithms) could improve different aspects of the evolutionary algorithms. According to the available research in the field of the evolutionary algorithm, covariance matrix adaptation evolutionary algorithm (CMA-ES) is more favored than the remaining evolutionary algorithms. This is due to the use of the Gaussian distribution for calculation of the evolution on one hand and performing the adaptation with its optimization parameters on the other. Thus, the process requires less user effort for adjustment of the initial parameters. In this study, we seek to use surrogate models in covariance matrix adaptation evolutionary strategy to use a successful initial surrogate model construct for adapting the optimization parameters of CMA-ES. Experimental validation shows that the proposed algorithm has advantages over the currently available algorithms due to the remarkable decrease of the function evaluation. Key Words evolutionary optimization, covariance matrix adaptation evolutionary strategy, machine learning , surrogate models