Table 3.

Performance comparison between a stand-alone classifier (CNN) and the proposed method.

Optimizer	Methods	Accuracy (%)	PPV (%)	Recall (%)	Specificity (%)	F1-score (%)	AUC	Loss	Total training time (s)
Adagrad	CNN	92.45	93.91	87.02	94.91	90.89	0.91	0.52	464.75
	Proposed method	98.78	100	98.16	99.42	99.00	0.99	–	464.75
RMSProp	CNN	93.48	94.56	89.99	95.13	91.03	0.93	0.48	471.22
	Proposed method	98.99	100	98.65	99.49	99.50	0.99	–	471.22
Adam	CNN	93.92	95.01	90.09	95.89	92.22	0.95	0.41	476.85
	Proposed method	99.18	100	98.88	99.66	99.70	0.99	–	476.85