Chapter 9

Question 1

What is a tree leaf node?

Answer 1

In a classification tree each time there is a partition, the measurements are divided into 2 leaf nodes.

Question 2

How to determine which questions is better in decision trees?

Answer 2

1) balance - how much each questions splits up a group.

2) Purity - we want the groups toonly contain one group (e.g. type of animal)

Question 3

What is impurity?

Answer 3

A number that represents how impure the root is and also how impure it is after split.

Question 4

Name impurity functions:

Answer 4

1) Entropy
2) Gini
3) ClassError

Question 5

What is early stopping?

Answer 5

Instead of stopping at totally pure splits, stops at different criteria:

1) when branch contains fewer than x observations
2) Stop at certain tree depth
3) when purity gain of best split below a certain value

Question 6

What is pruning?

Answer 6

Cutting branches after stopped algorhitm.