Table 5. Classifier performance estimates obtained from the 20-fold cross-validation scheme.

Fold #	gamma	coef0	cost	Sensitivity	Specificity	Accuracy	p-value
1	0.00010	0.12	150	1.00	1.00	1.00	0.0077
2	0.00060	0.09	150	1.00	0.86	0.93	0.0009
3	0.00005	0.13	175	1.00	0.50	0.67	0.6503
4	0.00005	0.09	175	1.00	0.75	0.88	0.0021
5	0.00005	0.15	185	1.00	0.80	0.92	0.0166
6	0.00005	0.12	160	1.00	0.67	0.83	0.1094
7	0.00005	0.09	160	1.00	0.40	0.67	0.3743
8	0.00005	0.09	185	1.00	0.44	0.58	0.9456
9	0.00005	0.08	190	0.70	0.80	0.73	0.4041
10	0.00005	0.12	180	1.00	0.86	0.94	0.0016
11	0.00005	0.13	180	1.00	0.50	0.63	0.8862
12	0.00005	0.40	240	1.00	1.00	1.00	0.0016
13	0.00005	0.15	120	0.83	1.00	0.88	0.3671
14	0.00005	0.20	170	1.00	0.60	0.80	0.0547
15	0.00020	0.90	240	1.00	0.86	0.92	0.0039
16	0.00005	0.40	120	1.00	1.00	1.00	0.0050
17	0.00005	0.20	90	1.00	0.80	0.89	0.0413
18	0.00030	1.10	240	1.00	0.83	0.93	0.0046
19	0.00050	0.90	200	1.00	1.00	1.00	0.0199
20	0.00005	0.30	120	1.00	0.83	0.92	0.0039

The tuned parameters (gamma, coef0, cost) and the performance estimates (sensitivity, specificity, accuracy) for each iteration are shown. The parameters corresponding to the best performance are shaded. A p-value from McNemar's Chi-square test was computed, and p < 0.05 was considered statistically significant.