From: Injury narrative text classification using factorization model
Measure | Definition | Equation |
---|---|---|
Positive Predictive Value (PPV) | Ratio of truth positive to number of cases truly falls under the classes | TP/(TP+FP) |
Sensitivity | Proportion of actual positives which are correctly identified | TP/(TP+FN) |
Specificity | Proportion of negative which are correctly identified | TN/(FP+TN) |
10 Cross Validation Accuracy | Split the dataset into 10 folds and runs classification 10 times | (TP+TN)/(TP+FP+TN+FN) |
Kappa | Robust measure taking into account the agreement by chance | K = (I o -I e )/(1-I e ) I o = (TP+TN)/(TP+TN+FP+FN) I e = ((TN+FN)(TN+FP)+ (TP+FP)(TP+FN))/n2 n = TP+TN+FP+FN |
Area Under Receiver Operating Characteristic Curve (AUC) | Identify the relation between sensitivity and specificity | Area under curve plotting sensitivity and (1-specificity) |