Journal of Advances in Mathematics and Computer Science

The essence of surrogate model is a low-cost alternative, which mainly replaces the computationally heavy simulation process to reduce the time cost consumed. In the past two decades, this approximation based optimization method has made remarkable progress, and surrogate models are widely used in computationally expensive simulation model analysis and optimization. In addition, with the development of technology, the surrogate model is no longer a simple substitute, but can drive new sample points to join the training process based on historical data, so as to gradually approach the global optimal solution of the problem. For optimization problems, there are many surrogates-assisted optimization algorithm methods. However, the selection of sample points has great influence on the accuracy of the surrogate model. In order to obtain a more accurate surrogate model, the newly added sample points should meet the sample diversity criterion of the specified distance, and at the same time, corresponding strategies should be adopted to fully explore sparse regions, so as to avoid falling into the local optimal phenomenon in the optimization process. Therefore, an ensemble of surrogates based on alternate point-taking strategy (APTS) is proposed, and a hierarchical search framework is designed, using different algorithms at each stage. The effectiveness of APTS is verified on three benchmark examples with different dimensions and compared with several advanced methods. The results show that this method has better accuracy and robustness than other methods on most test problems.