Class imbalance threshold number in classification problem

Is there a threshold to define class imbalance?

I currently have 75:25 of target 0 and 1, but I have found conflicting sources that imbalance starts from 80 or 90 above. Since I need to justify my model choice if imbalance is present, can anyone help to weight in or have more insight on this?

I’m currently doing ensemble, xgboost model for this project and use scale_pos_weight but the prediction does not change much.

Leave a Comment