Precision and Recall
Precision = TP/(TP+FP) measures the fraction of predicted positives that are true positives; recall = TP/(TP+FN) measures the fraction of true positives that are detected.
Purpose
To evaluate the two sides of classifier quality: exactness and coverage.
Mechanism
Precision: the fraction of predicted positives that are truly positive (low FP). Recall: the fraction of true positives that are predicted positive (low FN). The two usually trade off and must be balanced per task. The PR curve shows the tradeoff across thresholds.
Use cases
- Classification evaluation
- Information retrieval
- Medical diagnosis