TABLE 4:
Classifiers are retrained on filtered instances of ACES dataset and tested on NKI. The AUC gain of the classifiers on NKI dataset after filtering ACES instances for PCI, LIME and SHAP.
AUC for all ACES instances | AUC for filtered instances | |||
---|---|---|---|---|
Method | PCI | LIME | SHAP | |
RF | 0.732 | 0.738 | 0.721 | 0.734 |
LR | 0.730 | 0.731 | 0.731 | 0.735 |
lSVM | 0.732 | 0.736 | 0.735 | 0.736 |
rSVM | 0.715 | 0.716 | 0.714 | 0.720 |