0
$\begingroup$

I have total 300 categorical features and one response variable which is continuous. Its completely new to me. How can I select relevant categorical features for linear regression or multiple regression? What is the feature selection criteria and right algorithm for this type of problem?

Insights are much appreciated.

Improve this question
$\endgroup$
1
  • 1
    $\begingroup$ What do you want to do with the model? $\endgroup$
    – Matthew Drury
    Commented Jun 19, 2017 at 14:01

1 Answer 1

Reset to default
1
$\begingroup$

The most basic and intuitive type of variable selection algorithm is forward and backward stepwise selection. There are serious problems with using them, but they still get used quite often.

Another method that has become quite popular is using penalized regression, particularly the LASSO. This will automatically shrink some variables to zero allowing you to just keep the non-zero ones in your model.

Both of these methods are explained pretty well and accesibly in chapter 6 of An Introduction to Statistical Learning.

http://www-bcf.usc.edu/~gareth/ISL/ISLR%20Seventh%20Printing.pdf

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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