RSS

Precision/Recall/Accuracy

機械学習でよく出てきてよく混乱するアレです。

  •                 |   Positive (Actual)  | Negative (Actual)

———————–|———————-|———————– Positive (Prediction) | True Positive (TP) | False Positive (FP)
Negative (Prediction) | False Negative (FN) | True Negative (TN)


  • Precision (適合率、精度): Of examples recognized as positive, what percentages are actually positive? $$ {\rm Precision} = \frac{\rm TP}{{\rm TP} + {\rm FP}} $$

  • Recall (再現率): What percentages of actual positives are correctly recognized as positive? $$ {\rm Recall} = \frac{\rm TP}{{\rm TP} + {\rm FN}} $$

  • Accuracy (正解率): $$ {\rm Accuracy} = \frac{\rm TP}{{\rm TP} + {\rm TN}} $$