6

Question 1

What is the approximate depth of a decision tree trained with 1 million instances

Answer 1

Depth is calculated using log2(m) which roughly equals 20

Question 2

Is a nodes Gini greater or lower then its parents

Answer 2

Its generally lower as the CART training equation aims to reduce impurity

Question 3

If a decision tree is over fitting what should be done with the max_depth hyper parameter

Answer 3

Decrease max depth to constrain and regularize the model

Question 4

If a decision tree is under fitting should the input features be scaled

Answer 4

It makes no difference

Question 5

If it takes an hour to train a decision tree with 1 million instances how long will it take to train one with 10 million instances

Answer 5

11.7 hours

Question 6

How does presort affect training speed

Answer 6

It speeds up training if the number of instances is less then a few thousand any more and it will slow it down