Variable selection from a feature representing protein sequences: a case of classification on bacterial type IV secreted effectors

Table 1 Quantitative results by incrementally adding the variables after the 25th one of 400 variables in a descending order according to their accumulated scores and the comparison results shown in reference [6]

Measure	Max value	Mean value	Min value	Results in [6]		NB	kNN	LR	RF	ERT	SVM	XGB	GBM
Acc	0.910	0.906	0.893	Training set	1st stage	0.732	0.855	0.879	0.885	0.894	0.902	0.901	0.905
				Training set	2nd stage	0.909	0.910	0.911	0.904	0.889	0.906	0.907	0.906
				Testing set	2nd stage	0.938	0.935	0.944	0.938	0.940	0.945	0.924	0.931
AUC	0.940	0.932	0.918	Training set	1st stage	0.812	0.907	0.906	0.921	0.925	0.935	0.927	0.929
AUC	0.940	0.932	0.918	Training set	2nd stage	0.927	0.925	0.929	0.906	0.812	0.935	0.921	0.907

ISSN: 1471-2105