Table 9.

Comparisons of the proposed model at different batch sizes and learning rates

Sampling Technique	Batch Size	Learning Rate	Accuracy	Precision	Recall	F1-score
		0.001	97%	97%	95%	94%
	16	0.0001	95%	94%	97%	94%
		0.00001	95%	92%	95%	92%
		0.001	97%	93%	97%	96%
Stratified	32	0.0001	96%	88%	96%	90%
Sampling		0.00001	95%	92%	95%	94%
		0.001	97%	83%	97%	90%
	64	0.0001	97%	83%	97%	90%
		0.00001	94%	82%	94%	85%
		0.001	89%	64%	89%	75%
	16	0.0001	89%	64%	89%	75%
		0.00001	89%	64%	89%	75%
		0.001	88%	74%	88%	76%
Random	32	0.0001	89%	64%	89%	75%
Sampling		0.00001	89%	64%	89%	75%
		0.001	90%	92%	80%	72%
	64	0.0001	85%	84%	85%	77%
		0.00001	89%	63%	87%	70%
		0.001	94%	95%	94%	93%
	16	0.0001	96%	98%	96%	95%
		0.00001	88%	68%	88%	83%
		0.001	89%	88%	89%	87%
Holdout	32	0.0001	89%	88%	86%	87%
Sampling		0.00001	89%	88%	89%	87%
		0.001	95%	84%	95%	80%
	64	0.0001	89%	83%	84%	85%
		0.00001	88%	81%	88%	85%