Loading…
Glossary term
Of the items a model flagged as positive, the share that truly were positive.
Precision asks: of everything the model flagged as positive, how much really was? It equals true positives divided by true positives plus false positives. High precision means few false alarms.
| Predicted positive | Predicted negative | |
|---|---|---|
| Actually positive | True positive | False negative |
| Actually negative | False positive | True negative |
Precision
precision = tp / (tp + fp)Precision matters when false alarms are costly, like flagging innocent transactions as fraud.