I tried using MRMR on a dataset that about 10% of the dataset has class '1' and the remaining 90% has class '0'. I used the MRMR code shown below with K=10. However, I realized that after using count ifs there were more rows that each selected feature and the class had 0,0 than 1,1. Making me realize that the MRMR that the selected features might be predicting the absence of such features and not the presence of such features to the target class. Is there a way to use MRMR for imbalanced data?
use mrmr classification
selected_features = mrmr_classif(X, y, K = 10)
I would like the MRMR to select features that contribute to having class '1' and not class '0'
