Model Evaluation

---

Precision

• aka, positive predictive value --

\[\frac{TruePositives}{TruePositives + FalsePositives}\]

• It depends upon the prevalence, or, the proportion of samples that are, in fact, positive.

Recall (Sensitivity)

• aka, Sensitivity

\[\frac{TruePositives}{TruePositives + FalseNegatives}\]

• This is the probability of a positive result given that the instance is, in fact, positive.

F1 Score

• The harmonic mean of precision and recall, the F-measure:

\[F = 2 \cdot \frac{precision \cdot recall}{precision + recall}\]

---

• True positive: Sick people correctly identified as sick
• False positive: Healthy people incorrectly identified as sick
• True negative: Healthy people correctly identified as healthy
• False negative: Sick people incorrectly identified as healthy

Relationship of these metrics

---

              precision    recall  f1-score   support

           0       0.85      0.93      0.89      4687
           1       0.63      0.42      0.50      1313

    accuracy                           0.82      6000
   macro avg       0.74      0.67      0.69      6000
weighted avg       0.80      0.82      0.80      6000

• The precision of class 0 : \(0.85\)
• The recall (sensitivity) of class 0 : \(0.93\)

• The false discovery rate of class 0 : \(1 - 0.85 = 0.15\)

• The precision of class 1, the minority class : \(0.63\)

• The recall of class 1 : \(0.42\)
• The false discovery rate of class 1 : \(1 - 0.63 = 0.37\)

---