simple question:
does training with oversampling increase overfitting?
yes?
no?
depends on indicators?