Figure 6.

ROC curves and importance scores of the top five features ranked by average value in the margin classification task. (A1, A2, A3) ROC for each model in the Mar1, Mar2, and Mar3 classifications. (B1, B2, B3) The feature importance scores for the Mar1, Mar2, and Mar3 classifications. * denotes that the average values of PIS, LC, and MIS for that feature are greater than 0.75. ROC, receiver operating characteristic; AUC, area under the ROC curve; PIS, permutation importance score; LC, LASSO coefficient; MIS, mutual information score; L, low; H, high; DV, dependence variance; LASSO, least absolute shrinkage and selection operator.