0
$\begingroup$

I created a regression tree which predicts wheter people will buy an product or not. The tree is very accurate, but it is also very large and has a depth of 200.

When you want to learn something from the tree, you cant look at the whole tree, just because it is too large. So, my idea is to limit the depth of the tree, e.g. to 10.

But here is he problem. How can the depth of the tree be reduced? There are two options:

1. Create the whole tree with a depth of 200. An then just cut it at level 10.

2. Train a new tree and limit the size while training/ parameter optimization to 10.

Which option should be preferred?

Another problem is that the smaller tree is not as good as the large tree. Therefore, you might learn "bad" things from it.

Thanks!

$\endgroup$
  • $\begingroup$ If you cut a tree with 200 levels at 10, I'm reasonably certain your accuracy will go down substantially. I would train with the depth constraint. $\endgroup$ – KirkD_CO Jan 12 at 2:54

Your Answer

By clicking "Post Your Answer", you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.