Table 7.

Comparison of the CNN and the CNN-AE using different evaluation metrics based on a tenfold cross-validation.

Methods	Fold No	Accuracy (%)	PPV (%)	Recall (%)	Specificity (%)	F1-score (%)	AUC (%)	Loss	Total training time (s)
CNN	1	93.75	94	100	0	97	50	0.2157	27.04
	2	93.75	94	100	0	97	50	0.2104	20.48
	3	93.75	94	100	0	97	50	0.2733	21.34
	4	90.63	97	94	0	95	46.77	0.6140	21.49
	5	96.88	97	100	66.67	98	83.33	0.0916	21.70
	6	90.63	97	94	0	95	46.77	0.2164	22.00
	7	93.75	94	100	0	97	50	0.1642	21.51
	8	93.75	94	100	0	97	50	0.1816	21.83
	9	81.25	96	81	80	88	80.74	0.7327	21.97
	10	96.77	97	100	0	98	50	0.1214	22.75
95% confidence interval over 10 folds		92.49 ± 2.75	95.40 ± 0.88	96.90 ± 3.73	14.67 ± 19.00	95.90 ± 1.82	55.76 ± 8.54	0.282 ± 0.13	22.21 ± 0.37
CNN-AE	1	98.08	97	100	94.74	99	97.37	0.0925	33.01
	2	94.23	94	97	90.91	95	93.79	0.2600	31.50
	3	100	100	100	100	100	100	0.0096	31.35
	4	96.15	96	96	95.83	96	96.13	0.2600	31.60
	5	93.27	91	97	85	94	90.94	0.4017	31.90
	6	92.31	94	94	90.48	94	92.01	0.3678	31.99
	7	98.08	97	100	95.45	98	97.73	0.0858	32.30
	8	96.15	94	100	90.91	97	95.45	0.2027	32.90
	9	94.23	92	96	92.59	94	94.3	0.1572	33.13
	10	98.04	97	100	95.45	98	97.73	0.0614	33.83
95% confidence interval over 10 folds		96.05 ± 1.48	95.2 ± 1.63	98 ± 1.33	93.14 ± 2.52	96.5 ± 1.27	95.55 ± 1.70	0.19 ± 0.07	32.35 ± 0.49