F2-Score
The F2-Score is a weighted average of precision and recall: \[ F_2 = 5 \cdot \frac{precision \cdot recall}{4\cdot precision + recall} \]
Where
Precision is defined as \[ \frac{TP}{TP+FP} \] and
Recall is defined as \[ \frac{TP}{TP+FN}
\]
The F2-Score originates from the binary classification background, where we only have two classes that we want to distinguish:
positive and
negative.
In this scenario there are four possible outcomes:
- TP (True Positive): The object belongs to class positive and we classified it as positive,
- FP (False Positive ): The object belongs to class negative and we classified it as positive,
- TN (True Negative): The object belongs to class negative and we classified it as negative,
- FN (False Negative): The object belongs to class positive but we classified it as negative
| | Reality |
| | Positive | Negative |
Prediction | Positive | TP | FP |
Negative | FN | TN |