Table 4.

The base models of ANN and their evaluations.

ANN models	Training algorithm	Number of total hidden nodes	Hidden activation function	Output activation function	Architecture	Training					Testing
						R2	RMSE	MAE	VAF	Accuracy	R2	RMSE	MAE	VAF	Accuracy
ANN1	TrainSCG	4	Tansig	Tansig	7-4-1	0.934	0.660	0.428	93.415	91.471	0.724	1.193	0.724	68.725	87.644
ANN2	TrainSCG	7	Logsig	Tansig	7-7-1	0.937	0.693	0.283	93.100	94.829	0.643	1.459	0.756	57.458	85.260
ANN3	TrainLM	10	Tansig	Tansig	7-4-6-1	0.948	0.567	0.350	94.767	94.247	0.928	0.293	0.487	92.773	90.254
ANN4	TrainLM	12	Purelin	Tansig	7-5-7-1	0.883	0.864	0.535	87.290	89.503	0.802	1.395	0.820	77.386	85.371
ANN5	TrainOSS	13	Logsig	Logsig	7-5-8-1	0.932	0.672	0.411	93.213	91.816	0.850	0.492	0.508	84.935	90.061
ANN6	TrainGDX	14	Tansig	Logsig	7-7-7-1	0.939	0.684	0.483	93.774	91.529	0.906	0.754	0.666	89.952	87.247
ANN7	TrainLM	16	Logsig	Logsig	7-7-9-1	0.930	0.643	0.332	92.783	94.510	0.924	0.589	0.360	96.392	90.164
ANN8	TrainGDX	14	Purelin	Purelin	7-9-5-1	0.906	0.799	0.499	90.543	91.588	0.841	0.850	0.622	83.640	87.034
ANN9	TrainSCG	17	Tansig	Purelin	7-9-8-1	0.947	0.677	0.432	94.696	91.873	0.816	0.993	0.606	80.912	88.465
ANN10	TrainGDX	24	Logsig	Logsig	7-11-13-1	0.915	0.913	0.624	88.188	88.413	0.879	1.126	0.985	80.926	79.481
ANN11	TrainSCG	26	Tansig	Tansig	7-11-15-1	0.938	0.619	0.336	93.654	93.553	0.882	1.023	0.586	88.018	90.555
ANN12	TrainGDX	32	Purelin	Tansig	7-15-17-1	0.922	0.680	0.387	92.201	93.114	0.866	0.978	0.392	86.559	88.953
ANN13	TrainLM	37	Tansig	Tansig	7-17-20-1	0.906	0.900	0.628	88.409	87.744	0.763	0.992	0.822	73.105	85.883
ANN14	TrainSCG	39	Purelin	Logsig	7-17-22-1	0.855	0.916	1.501	87.608	64.877	0.765	0.596	1.357	79.748	87.901
ANN15	TrainLM	42	Tansig	Logsig	7-17-25-1	0.913	1.035	0.844	90.418	87.033	0.758	0.981	0.893	75.765	82.328

LM Levenberg–Marquardt, GDX Adaptive learning rate, SCG Scaled conjugate gradient, OSS One-step secant.