変数間の相互作用を考慮した変数削減法の提案

抄録

Reduction of the number of design variables is one of major issues in surrogate modeling and various methods are being proposed. In this paper, a method based on statistical analysis and can work with small size sample is argued. Three different DOE samples are considered and examined for constructing variable reduction model. It is concluded that adequate treatment of interaction effects among design variables is essential and hence resolution IV DOE is required for modeling samples.