Table 5.

Top results obtained for MDDSI classification using KNN and SVM-RBF classifiers with single physiological signal.

Classifier	Signal	Parameters	No. of features	Feature subset	ACC	PPV	SEN	SPC
KNN	ECG	K = 10	2	1 to 2	0.6932	0.7955	0.8139	0.8000
	PPG	K = 5	5	1 to 5	0.6136	0.8000	0.6512	0.8444
	RSP	K = 10	3	1 to 3	0.5795	0.6944	0.5814	0.7556
SVM-RBF	ECG	σ = 1.3	3	1 to 3	0.7273	0.8181	0.8372	0.8222
	PPG	σ = 0.5	2	1 to 2	0.5341	0.6604	0.8139	0.6000
	RSP	σ = 1.9	3	1 to 3	0.6136	0.6875	0.7674	0.6667

K is the number of nearest neighbors. σ is the bandwidth of kernel function. Number of ranked features used and their selected subset are according to Table 3.