Reading a Confusion Matrix

The four cells

For binary classification:

• TP — predicted positive, actually positive.
• TN — predicted negative, actually negative.
• FP — predicted positive, actually negative (false alarm).
• FN — predicted negative, actually positive (miss).

Derived metrics

• Precision = TP / (TP + FP) — trust of positive predictions.
• Recall = TP / (TP + FN) — coverage of real positives.

Why look at it

Accuracy hides which mistakes you make. The matrix shows whether your model errs toward false alarms or misses — and which one hurts more in your context.