Table 3.

Classifier properties

	ME	M.T.	Pat 1	Pat. 2	Pat. 3
AUC	0.809	0.868	0.657	0.679	0.733
Fraction of Negatives kept when loosing 50% of positives	0.101	0.051	0.322	0.000	0.246
Fraction of positives kept when loosing 50% of negatives	0.865	0.914	0.714	0.660	1.000
Accuracy	0.753	0.810	0.662	0.830	0.831
F1	0.391	0.331	0.207	0.795	0.214

For each dataset we provide, the AUC and F1 value as well as the max accuracy on the test set. Similarly, we provide the fraction of negatives maintained when keeping 50% of negative and the fraction of positives when removing 50% of negatives