To do this part of the process, we will build a model with the earth package, review it on the training data, then see how it performs on the test data. We'll run a 10-fold cross-validation with the algorithm:
> set.seed(1492)> earth_fit <- earth::earth( x = pca_scores[, 1:5], y = pca_scores[, 6], pmethod = 'cv', nfold = 10, degree = 1, minspan = -1 )
Calling the summary of the model object gives us seven total terms with three of the features:
> summary(earth_fit)Call: earth(x=pca_scores[,1:5], y=pca_scores[,6], pmethod="cv", degree=1, nfold=10, minspan=-1) coefficients(Intercept) 174.182h(0.1-PC1) -26.380h(PC1-0.1) 33.806h(0.01-PC2) -13.181h(PC2-0.01) 13.842h(0.02-PC5) 1.333h(PC5-0.02) -0.869Selected 7 of 7 terms, ...
