Confusion Matrix and ROC Curves

Let's explore two powerful tools for understanding classification model performance.

Reading a Confusion Matrix

Adjusting the Threshold

ROC Curve Intuition

AUC: Area Under the Curve

Multi-class Confusion Matrix

Key Takeaways

• Confusion matrix shows all 4 outcomes: TP, TN, FP, FN
• Threshold trades off precision vs recall
• ROC curve shows TPR vs FPR at all thresholds
• AUC summarizes ROC in a single number (0.5 = random, 1.0 = perfect)
• For multi-class: calculate metrics per class or use macro/micro averaging
• Use sklearn: confusion_matrix(), roc_curve(), roc_auc_score()