Concept - F1 Score

$F1=\frac{2\times \text{precision}\times \text{recall}}{\text{precision}+\text{recall}}$ (s.280)

• Conveys the balance between precision and recall (s.280).
• Motivation: accuracy does not perform well for imbalanced data sets. Example: 100 transactions, 90 non-fraud, 10 fraud; classifying everything as non-fraud gives high accuracy (90%) but catches no fraud. Precision and recall give more insight (s.280).

Related

Concept - Precision and Recall | Concept - Confusion Matrix

Source chunk: 03 Text Classification - Chunk 10 (s.271-300)