is there any rule to apply pca to the imbalance data? [closed]

Question

Is there any rule to apply PCA to imbalanced data? (randomforest, xgboost)

I used multiclass imbalance data to pca
but the log-loss accuracy getting decrease
any theoritical background of this?

Answer 1

PCA is method for unsupervised learning so you shouldn't add the label.

Imbalanced data refers to the label.

So for me it does not make too much sense to talk about PCA and imbalanced data.