Table 6. Proportion of test individuals for which the decision-path model is different from the path in CART model.

Dataset	DP-AUC	DP-IG	DP-BAY
pneumonia	0.31	0.17	0.23
sepsis-d	0.29	0.15	0.16
sepsis-s	0.24	0.18	0.20
heart failure-d	0.19	0.17	0.16
heart failure-c	0.26	0.18	0.19
HIT	0.21	0.12	0.14
Alzheimer	0.25	0.18	0.22
Mean	0.25	0.16	0.18

For each method the table gives the mean proportion obtained from 20-fold cross-validation. The last row gives the mean proportion over all datasets.