Table 5.

The prediction results by the five methods averaged over 13 classes

Methods	precision (%)	recall (%)	F1 (%)	ROC score	coverage a
AGPS	15	66	22	0.61	13 (13)
PSoL	20	18	19	0.55	12 (13)
two-class SVMs	28	10	16	0.53	11 (13)
two-class SVMs_balanced	18	36	29	0.57	10(13)
one-class SVMs	10	42	15	0.53	13 (13)
kernel integration	39	16	23	0.56	11(13)
kernel integration_balanced	11	32	24	0.59	6(13)

^a The digit in the parenthesis is the true number of functional classes whereas the number outside is the number of classes that can be predicted by the corresponding method.