The F-score is a metric used to evaluate the performance of a predictive system, particularly in classification tasks. It is the harmonic mean of precision and recall, calculated with the formula:
Where:
- Precision measures the accuracy of positive predictions (the proportion of true positives out of all predicted positives).
- Recall measures the system’s ability to identify all relevant positive instances (the proportion of true positives out of all actual positives).
The F1 score balances precision and recall, giving an overall measure of a model’s accuracy. However, a key criticism of the F-score is that a moderately high value may result from an imbalance between precision and recall, which doesn’t provide a complete picture of a system’s performance. For example, a model with high precision but low recall could have a good F1 score, but fail to detect many true positives.
In critical applications, where the cost of missing true positives is higher than the cost of false positives, a variant called the F2 measure can be used. This gives more weight to recall than precision. Conversely, when precision is more critical, the F0.5 measure can be used to emphasize precision over recall.
These variations allow for tailored evaluation based on the specific needs of the application, balancing the trade-offs between precision and recall depending on the importance of each.