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

Table 5

Results for the FsBoost.V1 ensemble algorithm (Levels 1 and 2).

Level	Level 1									Level 2
Classifier	kNN ensemble			PNN ensemble			SVM ensemble			Ensemble
Class	Sen	Spe	Acc	Sen	Spe	Acc	Sen	Spe	Acc	Sen	Spe	Acc
For A dataset
E	0.88	1.00	93.87	0.89	1.00	94.26	0.99	1.00	99.48	0.93	1.00	96.43
N-E	1.00	0.88		1.00	0.89		1.00	0.99		1.00	0.93
AUC	0.94			0.94			0.99			0.96
Kappa	0.88			0.89			0.99			0.93
F-measure	0.93			0.94			0.99			0.96
For B dataset
E	0.85	1.00	92.48	0.76	1.00	87.83	0.96	0.99	97.48	0.87	1.00	93.52
N-E	1.00	0.85		1.00	0.76		0.99	0.96		1.00	0.87
AUC	0.92			0.88			0.97			0.94
Kappa	0.85			0.76			0.95			0.87
F-measure	0.92			0.86			0.97			0.93
For C dataset
E	0.84	0.99	91.91	0.80	0.99	89.17	0.97	0.98	97.43	0.88	0.99	93.78
N-E	0.99	0.84		0.99	0.80		0.98	0.97		0.99	0.88
AUC	0.92			0.89			0.97			0.94
Kappa	0.84			0.78			0.95			0.88
F-measure	0.91			0.88			0.97			0.93
For D dataset
E	0.83	0.97	89.87	0.64	0.96	80.13	0.95	0.93	94.04	0.86	0.96	91.30
N-E	0.97	0.83		0.96	0.64		0.93	0.95		0.96	0.86
AUC	0.90			0.80			0.94			0.91
Kappa	0.80			0.60			0.88			0.83
F-measure	0.89			0.77			0.94			0.91

Sen: sensitivity, Spe: specificity, Acc: accuracy (%), E: epilepsy, and N-E: nonepilepsy.