Machine Learning · Chapter 27 of 40

ROC Curve & AUC

The ROC CURVE plots True Positive Rate vs False Positive Rate across all thresholds. AUC is the area under it — one number summarising ranking quality.

AUC 0.5 = random. AUC 1.0 = perfect.

Example 1 (python)

from sklearn.metrics import roc_auc_score
proba = model.predict_proba(X_test)[:, 1]
print(roc_auc_score(y_test, proba))

Output

0.92

Feed probabilities, not predictions.

Example 2 (python)

# For heavily imbalanced data, prefer PR AUC over ROC AUC

PR curve highlights the positive-class problem better.

Key points

ROC plots TPR vs FPR.
AUC 0.5 = random, 1.0 = perfect.
Uses predicted probabilities.
For imbalance, prefer PR AUC.

💡 Note: AUC is threshold-independent — great for comparing models, less useful when you must pick one threshold.

ROC Curve & AUC

Key points

📝 Quick Quiz