4. Metrics that Matter

1. Accuracy Paradox

You have a dataset of 99 healthy people and 1 sick person. Your model says "Healthy" for everyone.

• Accuracy: 99%.
• Usefulness: 0%.

2. Precision and Recall

• Precision: When I say it's Cancer, how often am I right? (Trust).
• Recall: Out of all Cancer patients, how many did I find? (Coverage).

python
from sklearn.metrics import classification_report
print(classification_report(y_true, y_pred))