SMOTE in Practice

SMOTE Example

from imblearn.over_sampling import SMOTE

# Create synthetic minority examples
smote = SMOTE(random_state=42)
X_train_balanced, y_train_balanced = smote.fit_resample(X_train, y_train)

# Now train on balanced data
model = LogisticRegression()
model.fit(X_train_balanced, y_train_balanced)

When to use each approach:

• Class weights: Simple, first try
• SMOTE: Better performance, but slower
• Threshold adjustment: Use alongside other methods
• Undersampling: Only if you have huge datasets and can afford to lose data