Open AccessThe ensemble of surrogate model based on local and global errors
Author(s) -
Xuejing Zhang,
H. Li,
Xiang Gao,
Hongyi Xu
Publication year - 2021
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/1043/5/052049
Subject(s) - surrogate model , kriging , robustness (evolution) , radial basis function , universality (dynamical systems) , benchmark (surveying) , computer science , mathematics , mathematical optimization , algorithm , artificial intelligence , machine learning , biochemistry , chemistry , physics , geodesy , quantum mechanics , artificial neural network , gene , geography
The weight coefficient is the key factor for the success of the ensemble of surrogate model construction. In this paper, a method for constructing ensemble of surrogate model by combining local and global errors to calculate weight coefficient has been put forward. Radial basis function (RBF) and the Kriging model are built as the meta models, and the cross validation strategy is applied to calculate the global errors and the local errors of samples. The inverse proportion average method is used to calculate the weight coefficient by combining the local errors and global errors. In order to verify the effectiveness of the proposed method, two meta models and three methods to construct the ensemble of surrogate models are tested with six benchmark functions. The results show that the proposed method can improve the accuracy, robustness and universality of the surrogate model.