A Novel Approach for Sleep Arousal Disorder Detection Based on the Interaction of Physiological Signals and Metaheuristic Learning

Table 3

The accuracy, AUC, and RR of the subset feature selection from the fractal descriptor are calculated under various scenarios. The results of SVM and improved SVM classifiers are compared to detect arousal events.

Signal type	No. of features	SVM			Optimized SVM
		ACC (%)	AUC (%)	RR (%)	ACC (%)	AUC (%)	RR (%)
EEG	16	74.88	75.64	73.11	83.31	84.19	83.21
	24	79.54	77.32	76.11	87.52	87.09	87.13
	32	74.14	74.37	75.21	85.18	85.54	84.53
	40	73.80	73.94	73.14	82.22	83.81	82.36
ECG	16	71.44	71.15	69.63	80.40	79.74	80.23
	24	74.10	74.41	73.30	82.29	82.13	81.53
	32	70.89	72.56	72.18	82.26	81.76	80.28
	40	68.37	67.51	66.30	81.87	80.54	81.47
EEG + ECG	16	78.29	78.35	78.58	89.74	89.04	89.28
	24	81.75	80.64	79.44	88.51	90.13	89.94
	32	78.67	77.18	78.43	88.47	88.03	87.27
	40	75.13	75.72	74.54	88.36	86.83	87.70

Bold values are the best values obtained.