Problem
1. Using the code presented in Section 8.6:
(a) Generate a dataset (X, y).
(b) Apply a PCA transformation on X, which we denote X?.
(c) Compute MDI, MDA, and SFI feature importance on (X?,y), where the base estimator is RF.
(d) Do the three methods agree on what features are important? Why?
2. From exercise, generate a new dataset (X¨,y), where X¨ is a feature union of X and X? .
(a) Compute MDI, MDA, and SFI feature importance on (X¨,y), where the base estimator is RF.
(b) Do the three methods agree on the important features? Why?