What is ensemble diagnostics?
The power of ensemble methods was demonstrated in the preceding chapters. An ensemble with decision trees forms a homogeneous ensemble, and this was the main topic of Chapter 3, Bagging, to Chapter 6, Boosting Refinements. In Chapter 1, Introduction to Ensemble Techniques, and Chapter 7, The General Ensemble Technique, we had a peek at stacked ensembles. A central assumption in an ensemble is that the models are independent of one another. However, this assumption is seldom true, and we know that the same data partition is used over and over again. This does not mean that ensembling is bad; we have every reason to use the ensembles while previewing the concerns in an ensemble application. Consequently, we need to see how close the base models are to each other and overall in their predictions. If the predictions are close to each other, then we might need those base models in the ensemble. Here, we will build logistic regression, Naïve Bayes, SVM, and...
