AI Blog

AI Blog

by Michele Laurelli

Precision and Recall

/prɪˈsɪʒən ənd rɪˈkɔːl/
Concept
Definition

Metrics for classification: Precision is correct positives / predicted positives; Recall is correct positives / actual positives.

Precision measures accuracy of positive predictions (avoiding false positives). Recall measures completeness (avoiding false negatives). The F1-score combines both into a single metric.

Examples

1

Medical diagnosis (high recall)

2

Spam detection (high precision)

3

Information retrieval

Michele Laurelli - AI Research & Engineering