Handling Imbalanced & Messy Data

Day 18 of 100 Days of MLOps. Fraud is rare, disease is rare, luxury homes are rare — and on imbalanced data a lazy model that predicts 'no' every time can score 90% accuracy while catching zero of the cases that matter. See the trap with real numbers, then fix it with the right metrics and class_weight='balanced'. Runs 100% locally on any OS.