05·Confusion matrix and ROC curve

Introduction

The confusion matrix is the foundation of classifier diagnostics — every other metric (precision, recall, specificity, F1) is derived from it. The ROC curve and its area (AUC) describe model behaviour across the FULL range of decision thresholds, not just one arbitrary 0.5 cutoff. This lesson breaks the confusion matrix down to its primitives, shows how to tune the threshold for business cost, when ROC lies (heavy class imbalance → prefer PR-AUC), and how to interpret AUC=0.5, AUC=1.0, and AUC<0.5.