Constraint Programming Based Biomarker Optimization

Table 1

Performance comparison of the algorithm fsCoP. fsCoP has no prefixed features, and the model fsCoP(ACE2) has two predetermined features.

GSE5406
fsCoP	Sn	Sp	Acc	Avc	MCC
SVM	1.000	1.000	1.000	1.000	1.000
NBayes	1.000	0.998	1.000	0.999	0.999
DTree	0.992	0.800	0.967	0.896	0.848
Lasso	0.999	0.900	0.987	0.950	0.939
KNN	1.000	0.871	0.983	0.936	0.923
fsCoP(ACE2)	Sn	Sp	Acc	Avc	MCC
SVM	1.000	0.996	0.999	0.998	0.998
NBayes	1.000	0.998	1.000	0.999	0.999
DTree	0.993	0.796	0.967	0.894	0.847
Lasso	1.000	0.907	0.988	0.953	0.944
KNN	0.999	0.860	0.982	0.930	0.916
GSE1869
fsCoP	Sn	Sp	Acc	Avc	MCC
SVM	1.000	0.955	0.983	0.978	0.965
NBayes	1.000	0.972	0.990	0.986	0.979
DTree	0.907	0.000	0.567	0.453	NaN
Lasso	0.960	0.989	0.971	0.974	0.943
KNN	1.000	0.994	0.998	0.997	0.996
fsCoP(ACE2)	Sn	Sp	Acc	Avc	MCC
SVM	1.000	0.939	0.977	0.970	0.953
NBayes	1.000	1.000	1.000	1.000	1.000
DTree	0.987	0.000	0.617	0.493	NaN
Lasso	0.990	0.967	0.981	0.978	0.962
KNN	1.000	1.000	1.000	1.000	1.000