Imbalanced Data Quick Reference

Quick reference for handling imbalanced data.

Resampling

# SMOTE
from imblearn.over_sampling import SMOTE
smote = SMOTE(random_state=42)
X_res, y_res = smote.fit_resample(X_train, y_train)

# Class weights
model = RandomForestClassifier(class_weight='balanced')

Metrics

# Use F1-score, PR-AUC for imbalanced data
from sklearn.metrics import f1_score, average_precision_score

Try next: Report PR-AUC or recall at a fixed precision on your next imbalanced task.