0
$\begingroup$

Let's assume that we have a pretty decent idea of which features are important, and let's assume that those features are not a lot, like 2 or 3, some categorical, some numerical.

Is there any reason to use random forest or boosting over other classification methods like SVM with regularization for cases like this?

I thought random forest or boosting are effective over other methods when there are a lot of features, and we don't know selecting which features would be optimal.

In cases like this in which the important features are known and there are not a lot of them (like 2~3), are there methods that generally outperform random forest or boosting?

$\endgroup$
0
$\begingroup$

Feature selection is a recommended approach before predictive modeling because the presence of correlated or irrelevant features may make the built model bad. But domain knowledge has its own importance. It is better to see if feature selection helps you create in a better model.

Feature selection in Python using scikit-learn provides a nice description of methods that you may use. RFECV are generally good technique in this regard.

$\endgroup$

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.