Table 3.

Performance comparison between our algorithm (‘SCI-BN’), SVM with the same set of features (‘SCI-SVM’), Clique based method using only the density feature (‘Density’) and the ‘MCODE’ methods (Bader et al., 2003b) (‘MCODE’)

Train	Test	Method	Precision	Recall	F1
MIPS	TAP06	Density	0.217	0.409	0.283
MIPS	TAP06	MCODE	0.293	0.088	0.135
MIPS	TAP06	SCI-SVM	0.247	0.377	0.298
MIPS	TAP06	SCI-BN	0.312	0.489	0.381
TAP06	MIPS	Density	0.143	0.515	0.224
TAP06	MIPS	MCODE	0.146	0.063	0.088
TAP06	MIPS	SCI-SVM	0.176	0.379	0.240
TAP06	MIPS	SCI-BN	0.219	0.537	0.312

Evaluation is based on precision, recall and the F1 measure. Experiments carried out with either MIPS as positive training set and TAP06 as test set, or vice versa.