Development and Validation of Embedded Device for Electrocardiogram Arrhythmia Empowered with Transfer Learning

Table 7

Class-wise training and validation results of the proposed DVEEA-TL model.

T	Evaluation matrix	F	N	V	Q	S
		Fusion beat (%)	Normal beat (%)	Ventricular beat (%)	Unknown beat (%)	Supraventricular beat (%)
Accuracy	Training	99.91	99.95	99.88	100	100
	Validation	99.81	99.78	99.82	100	100
Classification miss rate	Training	0.09	0.05	0.12	0	0
	Validation	0.19	0.22	0.18	0	0
Sensitivity	Training	99.56	99.74	100	100	100
	Validation	98.90	100	100	100	100
Specificity	Training	19.74	27.13	25.07	25.07	28.19
	Validation	20.03	34.93	23.69	25.37	28.55
Precision	Training	99.89	100	54.70	100	100
	Validation	100	100	99.04	100	100
FPR	Training	0.80	0.73	0.75	0.75	0.72
	Validation	0.80	0.65	0.76	0.75	0.71
FNR	Training	0.004	0.002	0	0	0.004
	Validation	0.01	0	0	0	0