I have created a gbm model in caret.
I wish to plot a learning curve with performance metric (AUC/ROC in my case) on y axis and 1:n.trees on x axis.
I want to get performance values on both train and test set for every tree created by gbm. Which is possible in gbm package (n.trees = 1:n.trees in predict.gbm()), but for some reason I can not reproduce the example in caret.
To simulate the whole train () process (including CV, etc.) with various subsets of the training set (10%, 20% ... 100%).In principle, I would like to evaluate whether your training set size will reduce the large variance.For 2. you need the error rate when changing the mtry hyperparameter, but if I understand what you mean, trainFit will calculate the error rate for different CV-folds (although, I suppose, the last figure is the average for everyone). do my homework
To simulate the whole train () process (including CV, etc.) with various subsets of the training set (10%, 20% ... 100%).In principle, I would like to evaluate whether your training set size will reduce the large variance.For 2. you need the error rate when changing the mtry hyperparameter, but if I understand what you mean, trainFit will calculate the error rate for different CV-folds (although, I suppose, the last figure is the average for everyone). do my homework
