Table 4.
Performance metrics for the adaptive synthetic (ADASYN) oversampled cross- validated lasso model for attacks on healthcare
| Performance metric | Healthcare infrastructure |
| cv_metric_AUC | 0.9467 |
| logLoss | 0.2963 |
| AUC | 0.9595 |
| prAUC | 0.9590 |
| Accuracy | 0.8976 |
| Kappa | 0.7952 |
| F1 | 0.8973 |
| Sensitivity | 0.9005 |
| Specificity | 0.8947 |
| Pos_Pred_Value | 0.8942 |
| Neg_Pred_Value | 0.9011 |
| Precision | 0.8942 |
| Recall | 0.9005 |
| Detection_Rate | 0.4475 |
| Balanced_Accuracy | 0.8976 |
To balance the samples, 1877 attacks against healthcare were compared against the original 1901 non-healthcare attacks. See the mikropml vignette for definitions of these metrics along with the R coding walkthrough (https://cran.r-project.org/web/packages/mikropml/vignettes/introduction.html).