14 - Classification & Regression Trees

Question 1

what are Decision Trees (or CARTS)

Answer 1

they partition the predictor space by simple binary splitting rules. Think ‘flow chart’

Question 2

What does an excessive amount of splitting lead to?

Answer 2

Overfitting

Question 3

Goal of growing a Regression Tree?

Answer 3

Predict the mean response for the observations that fall into each region Rj from that training set

Question 4

Goal of growing a Classification Tree?

Answer 4

Predict the categorical response as the most commonly occurring in. each region Rj from the training set

Question 5

How do we ‘best’ partition the predictor space for regression?

Answer 5

find binary splits which minimize the RSS