1
$\begingroup$

I am testing out different models for a regression task. When using OLS, Ridge and Lasso, I use different polynomial degrees of the explanatory variables. Example: For two variables x and y, degree 2 would give the explanatory variables x, x^2, xy, y, y^2.

When using decision tree, however, I am not sure whether it makes sense to use any any higher degrees than 1 as explanatory variables. Example: Does it make sense to test for x^2, xy and y^2 when applying a decision tree regressor?

The reason I ask, is that the decision tree regressor is a non-linear regressor. On the one hand this could perhaps be an arguement it not making sense to include higher order polynomials, as the decisoin tree already can deal with non-linearity.

$\endgroup$
1
$\begingroup$

Since CARTs (Classification And Regression Tree) are a non-parametric algorithm, they should be able to find interactions between variables and non-linear behaviors.

Nevertheless, building polynomials can help them have a better performance.

| improve this answer | |
$\endgroup$
  • $\begingroup$ Thanks for answering. I'll apply higher degree polynomials also for the decision tree regressor. $\endgroup$ – KJA Jan 21 at 6:51

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.