Multiple Infill Sampling Strategy using Multi-Surrogate Modelling for Global Optimization Problem

Abstract

In recent years, optimization using a surrogate model or metamodel received great scholarly attention in solving computer simulation problems. Surrogate models are fast approximation, high-fidelity models and better accuracy in the prediction of the model. The infill sampling strategy is the one-way method to refine the surrogate model and improve the accuracy of the model. This paper proposed a multiple adaptive sampling strategy using two (2) surrogate models, Radial Basis Function and Kriging. The proposed method of the multi-surrogate model predicts two sample points with a combination of DBSCAN clustering as the initial processing of training points. This approach helps to improve the performance of the algorithm in terms of accuracy by calculating root mean square error (RMSE). The contribution of an algorithm proposed 2 sample points prediction at one iteration instead of previous research only predicting one sample for each iteration. The algorithm was tested and demonstrated using low dimension test function benchmark from previous work.