A Novel Approach to Ensemble Classifiers: FsBoost-Based Subspace Method

Table 6

Level 1 classification results for the C dataset.

k-nearest neighbor classification algorithm
NP	k = 2			k = 2			k = 2
	DF
	Euclidean			Euclidean			Minkowski
FS	0			1			2
NF	68			25			15
Class	Sen	Spe	Acc	Sen	Spe	Acc	Sen	Spe	Acc
E	0.80	1.00	90.00	0.83	0.88	90.96	0.88	0.96	92.04
N-E	1.00	0.80		0.99	0.96		0.96	0.88
AUC	0.90			0.82			0.92
Kappa	0.80			0.90			0.84
F-measure	0.89			90.96			0.91
10-fold (%)	85.39			87.54			90.41
Probabilistic neural networks
NP	Spread			Spread			Spread
	0.11			0.21			0.21
FS	0			1			2
NF	68			25			15
Class	Sen	Spe	Acc	Sen	Spe	Acc	Sen	Spe	Acc
E	0.80	0.98	89.04	0.79	0.78	88.91	0.78	0.88	82.96
N-E	0.98	0.80		0.99	0.88		0.88	0.78
AUC	0.89			0.78			0.89
Kappa	0.78			0.88			0.78
F-measure	0.88			88.91			0.88
10-fold (%)	0.11			0.21			0.21
Support vector machines
NP	BoxConstraint
	4			88			4
FS	0			1			2
NF	68			25			15
Class	Sen	Spe	Acc	Sen	Spe	Acc	Sen	Spe	Acc
E	0.98	0.99	98.13	0.97	0.95	97.17	0.95	0.96	95.57
N-E	0.99	0.98		0.97	0.96		0.96	0.95
AUC	0.98			0.94			0.97
Kappa	0.96			0.97			0.95
F-measure	0.98			97.17			0.97
10-fold (%)	98.48			97.37			95.41

DF: distance function, Sen: sensitivity, Spe: specificity, Acc: accuracy (%), NP: network parameters, FS: feature selection, NF: number of features, EC: ensemble classifier, E: epilepsy, and N-E: nonepilepsy.