From: Supervised learning for infection risk inference using pathology data
Metric | Description | Equation |
---|---|---|
Sensitivity | Proportion of observed positives that are correctly identified as such (i.e. percentage of culture-positive profiles correctly identified as positive). Also called recall (REC) or true positive rate (TPR). | \(SENS=\frac {TP}{TP+FN}\) |
Specificity | Proportion of observed negatives that are correctly identified as such (i.e. percentage of culture-negative profiles correctly identified as negative). Also called true negative rate (TNR). | \(SPEC=\frac {TN}{TN+FP}\) |
ROC | This curve illustrates the performance of a binary classifier as its discrimnation threshold is varied by plotting true positive rate (TPR) against false positive rate (FPR). It is related to cost/benefit analysis of diagnostic decision making. | Â |
PR | This curve represents precision against recall where high scores for both shows that the classifier is returning accurate results (high precision) as well as returning a majority of all positive results (high recall). | Â |