Construction and Evaluation of Metamodels
K-fold Cross Validation
5D benchmark from nuclear criticality safety assessments
Metamodels
Setting up the Cross Validation
Construction of a formula Y~s(X1)+...+s(Xp)
Construction of a formula Y ~ X1+...+Xp
Mean Absolute Error
Comparison of different types of metamodels
Fitting metamodels
Prediction at newdata for a fitted metamodel
Choice of the penalty parameter for a PolyMARS model
Multiple R-Squared
Plot residuals
Relative Maximal Absolute Error
Root Mean Squared Error
Evolution of the stepwise model
Test the robustess of the cross-validation procedure
A set of test data
Estimation, validation and prediction of models of different types : linear models, additive models, MARS,PolyMARS and Kriging.