Table 2.

Sleep stage classification results.

Classifier Parameters		Classification Accuracy
k-Nearest Neighbors		Synchronization Likelihood		Relative Wavelet Entropy
k	Distance	Train (%)	Test (%)	Train (%)	Test (%)
1	Euclidean	100.00	72.55	100.00	89.95
3	Euclidean	85.80	74.86	96.26	90.90
5	Euclidean	83.40	76.22	94.39	91.44
1	Cityblock	100.00	67.93	100.00	88.18
3	Cityblock	84.40	71.33	95.79	89.40
5	Cityblock	80.48	73.10	94.16	89.95
1	Cosine	100.00	73.91	100.00	89.95
3	Cosine	88.66	77.17	96.26	91.85
5	Cosine	85.97	78.26	94.45	91.44
Support Vector Machines		Synchronization Likelihood		Relative Wavelet Entropy
C	Kernel	Train (%)	Test (%)	Train (%)	Test (%)
0.1	Linear	73.35	71.74	87.61	85.05
10	Linear	89.42	79.35	99.24	90.22
100	Linear	93.28	76.49	100.00	90.49
0.1	Polynomial, d = 3	53.83	52.72	99.65	91.44
10	Polynomial, d = 3	90.77	80.98	100.00	91.58
100	Polynomial, d = 3	99.59	82.07	100.00	91.58
0.1	Polynomial, d = 5	51.14	51.22	100.00	91.58
10	Polynomial, d = 5	86.44	78.94	100.00	91.58
100	Polynomial, d = 5	98.48	82.34	100.00	91.58
0.1	Gaussian, σSL = 1.95 σRWE = 0.25	71.65	67.53	80.30	78.13
10	Gaussian, σSL = 1.95 σRWE = 0.25	100.00	86.82	100.00	92.93
100	Gaussian, σSL = 1.95 σRWE = 0.25	100.00	86.82	100.00	92.80
0.1	Gaussian, σSL = 1.45 σRWE = 0.75	72.71	69.57	78.73	74.73
10	Gaussian, σSL = 1.45 σRWE = 0.75	100.00	86.28	100.00	92.66
100	Gaussian, σSL = 1.45 σRWE = 0.75	100.00	86.28	100.00	92.66
Neural Networks		Synchronization Likelihood		Relative Wavelet Entropy
Layer 1	Layer 2	Train (%)	Test (%)	Train (%)	Test (%)
10	–	85.80	78.26	94.16	87.77
30	–	83.82	79.48	96.67	90.49
50	–	84.04	78.94	95.79	90.22
100	–	86.62	80.16	94.39	88.59
50	50	86.73	78.13	96.20	89.54
100	50	86.09	79.89	97.49	89.13
100	100	86.97	79.08	95.27	88.72

Bold values are the maximum achieved accuracy for each feature extraction method.