Is there any basis in determining the probability threshold in xgb.predict_proba while using XGBClassifier, especially if the data are imbalanced. I know when xgb.predict is used, a probability threshold of 0.5 is assigned by default.
Is there any basis in determining the probability threshold in xgb.predict_proba while using XGBClassifier, especially if the data are imbalanced. I know when xgb.predict is used, a probability threshold of 0.5 is assigned by default.