Table 3.

Results obtained from the training and test sets that are given in Figure 7. The AUC could not be computed since the test set instances are all from a single class. Results in the first column are obtained by model averaging over all 3567 MBs.

Performance measure	ISMB complete model averaged	ISMB	ISMB-MS	NISMB
Misclassification error	0.0000	0.0000	0.3333	0.3333
AUC	-	-	-	-
Logarithmic loss	0.0406	0.0505	0.0596	0.0585
Squared error	0.0684	0.0783	0.0902	0.0862
CAL score	0.3720	0.4092	0.4534	0.4284