Table 1 Common metrics in model evaluation [28]
Evaluation metrics | Definitions | Formula |
|---|---|---|
Accuracy | Percentage of correct identification of both positive and negative results. The higher the accuracy, the better the classifier | \(\frac{\mathrm{TP}+\mathrm{TN}}{\mathrm{TP}+\mathrm{TN}+\mathrm{FP}+\mathrm{FN}}\) |
Precision/Positive Prediction Value (PPV) | Proportion of correctly identified positives out of all samples identified as positive | \(\frac{\mathrm{TP}}{\mathrm{TP}+\mathrm{FP}}\) |
Sensitivity/True Positive Rate (TPR) | Proportion of correctly identified positives | \(\frac{\mathrm{TP}}{\mathrm{TP}+\mathrm{FN}}\) |
Specificity/True Negative Rate (TNR) | Proportion of correctly identified negatives | \(\frac{\mathrm{TN}}{\mathrm{TN}+\mathrm{FP}}\) |
Negative Predictive Values (NPV) | Proportion of correctly identified negatives out of all samples identified as negative | \(\frac{\mathrm{TN}}{\mathrm{TN}+\mathrm{FN}}\) |
F-measure | Harmonic mean of precision and sensitivity. The highest F1 score is 1 and the lowest is 0 | \(\frac{2 *\mathrm{ TP}}{(2 *\mathrm{ TP }+\mathrm{ FP}+\mathrm{FN})}\) |