04·Classification metrics — accuracy, precision, recall, F1

Introduction

Accuracy alone is often misleading — especially under class imbalance or asymmetric error costs. This lesson dissects the metric zoo: precision, recall, F1, F-beta, specificity, balanced accuracy, MCC (Matthews Correlation Coefficient), Cohen's kappa, log-loss. We show when each metric fits a business problem — and why "high accuracy" in cancer detection or fraud can mean a useless model.