Cherrypicking one of many correlated variables

(Safouane Chergui) #1

Hello everyone! I’m in the phase of feature engineering (so haven’t got to modeling yet) and I have many correlated variables (some of them are very highly correlated). I would like to pick only one of the correlated group of variables. Is there a way to accomplish this?

I would use Lasso as stated in here https://www.quora.com/Given-several-highly-correlated-variables-how-can-I-pick-what-is-the-best-predictor-for-the-others but I’d like to get some other ideas.

Thanks

0 Likes