Table 12.

Performance comparison of models on original and augmented data (CNN, A-CNN, LSTM, A-LSTM, GRU, A-GRU).

Model	Data set	Parameters		Complexity	Assessment Metrics
		Optimizer	LR	SC	TC	Acc (%)	Sp (%)	Sn (%)	Pre (%)	F1-S (%)	AUC (%)
CNN	Original	SGD	0.0001	28.03M	5.05h	92.33	89.02	90.08	96.34	98.30	96.05
		ADAM	0.0001	28.01M	5.03h	92.84	93.30	95.34	97.00	97.89	97.93
CNN	Augmented	–	–	27.06M	4.07h	93.87	98.03	95.90	98.64	97.00	98.34
		-	–	27.07M	4.09h	93.98	99.10	97.34	98.89	98.33	98.90
A-CNN	Augmented	–	–	28.06M	5.02h	94.00	98.03	95.90	98.64	97.00	98.64
		–	–	28.08M	4.04h	95.78	98.77	98.94	99.09	99.34	99.00
LSTM	Original	–	–	39.02M	4.01h	96.30	90.50	92.28	98.98	98.00	98.84
		–	–	39.05M	9.08h	97.54	93.00	95.49	99.00	98.08	99.00
LSTM	Augmented	-	–	38.03M	5.08h	97.44	94.60	96.88	97.58	98.90	90.04
		–	–	39.02M	5.03h	97.54	95.77	97.01	98.66	98.98	99.70
A-LSTM	Augmented	–	–	37.01M	5.03h	98.00	97.82	97.18	87.00	99.30	99.04
		-	–	41.01M	5.04h	98.64	97.47	98.91	98.96	98.98	98.90
GRU	Original	–	–	110.01M	5.04h	97.90	98.09	99.00	98.34	99.00	98.97
		–	–	110.05M	5.04h	98.02	99.00	99.17	99.03	99.69	98.99
GRU	Augmented	–	–	113.05M	6.05h	97.92	96.39	94.09	94.87	98.58	98.06
		-	–	115.08M	6.04h	98.40	90.98	97.56	98.78	99.01	98.79
A-GRU	Original	–	–	121.07M	6.06h	98.44	99.59	97.60	99.03	98.30	97.95
		-	–	122.06M	7.09h	98.79	99.80	98.87	99.66	99.65	98.99
A-GRU	Augmented	–	–	122.03M	7.03h	98.97	99.35	98.89	99.99	99.96	99.66
		–	–	122.07M	7.08h	99.32	99.78	99.12	100	99.01	99.89

The number of trainable parameters in a model represents its space complexity. M stands for million. The more trainable parameters there are, the higher the space complexity (SC). The models’ training time (measured in hours) is known as the time complexity (TC).

Significant values are in [bold].