I have a training set of 260,000 observations and 30 IVs and with binary class imbalance 1:6 (yes, it does mess up models' performance), but using SMOTE isn't an option, since it takes forever on my laptop with this amount of data. Is there any better alternative to SMOTE in terms of computational speed and efficiency? I wanted to try something more effective than random under-sampling and class weights. Would be happy to hear any suggestions for use in R!
Asked
Active
Viewed 358 times
0
-
Questions to SO [tag:r] tag should be focused on a specific example including complete reproducible input, code and expected output. Generic questions are normally closed. – G. Grothendieck Mar 28 '21 at 13:45
-
@G.Grothendieck is it ok to leave this question here without the "r" tag then? – user000 Mar 28 '21 at 13:47
-
Maybe. I normally am focused on R. – G. Grothendieck Mar 28 '21 at 13:49