Journal of Big Data

Table 6 Confusion matrix of CNN-IKOA at various threshold levels

From: CNN-IKOA: convolutional neural network with improved Kepler optimization algorithm for image segmentation: experimental validation and numerical exploration

Predicted classes
	T-3					T-4
		Normal	COVID	Virus	Recall (%)	Normal	COVID	Virus	Recall (%)
Actual classes	Normal	129	4	1	96.26	130	1	3	97.01
	Covid	2	100	3	95.23	1	102	3	96.23
	Virus	10	2	112	90.32	6	1	117	94.35
Precision (%)		91.48	94.33	96.55	–	94.89	98.08	95.12	–

	T-5					T-7
		Normal	COVID	Virus	Recall (%)	Normal	COVID	Virus	Recall (%)
Actual classes	Normal	119	2	13	88.81	130	1	3	97.01
	Covid	1	103	1	98.09	2	102	1	96.23
	Virus	4	2	118	95.16	14	2	108	87.10
Precision (%)		91.48	96.26	96.55	–	89.04	97.14	95.58	–

	T-8					T-10
		Normal	COVID	Virus	Recall (%)	Normal	COVID	Virus	Recall (%)
Actual classes	Normal	127	4	3	94.78	127	2	5	94.78
	Covid	0	105	0	100	3	101	1	96.19
	Virus	4	3	117	94.35	8	3	113	91.13
Precision (%)		96.95	93.75	97.50	–	92.03	95.28	94.96	–

	T-12					T-15
		Normal	COVID	Virus	Recall (%)	Normal	COVID	Virus	Recall (%)
Actual classes	Normal	129	3	2	96.27	130	3	1	97.01
	Covid	0	105	0	100	0	105	0	100
	Virus	3	2	119	95.97	6	2	116	93.55
Precision (%)		97.73	95.45	98.35	–	95.59	95.45	98.31	–

The bold values highlight the best precision and recall values obtained for three classes (Normal, COVID, Virus)

Back to article page