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

Table 4

Level 1 classification results for the A dataset.

k-nearest neighbor classification algorithm
NP	k = 2			k = 2			k = 2
	DF
	Euclidean			Seuclidean			Euclidean
FS	0			1			2
NF	68			25			15
Class	Sen	Spe	Acc	Sen	Spe	Acc	Sen	Spe	Acc
E	0.84	1.00	92.17	0.87	0.89	93.26	0.89	1.00	94.57
N-E	1.00	0.84		1.00	1.00		1.00	0.89
AUC	0.92			0.87			0.94
Kappa	0.84			0.93			0.88
F-measure	0.92			93.26			0.93
10-fold (%)	88.85			91.22			92.54
Probabilistic neural networks
NP	Spread			Spread			Spread
	0.11			0.11			0.21
FS	0			1			2
NF	68			25			15
Class	Sen	Spe	Acc	Sen	Spe	Acc	Sen	Spe	Acc
E	0.87	1.00	93.26	0.94	0.75	93.09	0.75	1.00	87.70
N-E	1.00	0.87		0.92	1.00		1.00	0.75
AUC	0.93			0.86			0.94
Kappa	0.87			0.93			0.89
F-measure	0.93			93.09			0.94
10-fold (%)	0.11			0.11			0.21
Support vector machines
NP	BoxConstraint
	3			21			2
FS	0			1			2
NF	68			25			15
Class	Sen	Spe	Acc	Sen	Spe	Acc	Sen	Spe	Acc
E	0.99	1.00	99.65	0.99	0.98	99.39	0.98	0.99	98.83
N-E	1.00	0.99		1.00	0.99		0.99	0.98
AUC	1.00			0.99			0.99
Kappa	0.99			0.99			0.99
F-measure	1.00			99.39			0.99
10-fold (%)	99.76			99.54			99.02

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.