Table 2. Analysis of the Performance for Activation Functions in the Hidden Layers of the ANN Model.

function	equation	range	set	R2	RMSE	AARD/%	SDav
hyperbolic tangent (tansig)		[−1, 1]	train	0.9998	0.4299	3.9778	0.4268
			validate	0.9991	1.0437	4.3304	1.0434
			test	0.9994	0.8629	4.5404	0.8627
			total	0.9997	0.5847	4.0694	0.5827
logistic sigmoid (logsig)		[0, 1]	train	0.9998	0.4268	3.8884	0.4244
			validate	0.9990	1.1260	3.8893	1.1250
			test	0.9987	1.1822	6.1410	1.1800
			total	0.9996	0.6651	4.1140	0.6639
linear (purelin)		[−∞, ∞]	train	0.9011	11.3709	92.1066	11.3732
			validate	0.8988	11.5049	87.1327	11.4848
			test	0.8785	12.6966	76.6944	12.7037
			total	0.8984	11.5324	90.0660	11.5336
radial basis (radbas)		[0, 1] | x ∈ [0, ∞)	train	0.9691	5.5479	47.5140	5.5477
			validate	0.9425	7.8847	45.1713	7.8874
			test	0.9496	7.1821	57.1790	7.1870
			total	0.9643	6.0473	48.2469	6.0469
rectified linear unit (ReLU)		[0, ∞)	train	0.9982	1.4754	14.1345	1.4747
			validate	0.9948	2.4795	14.8493	2.4790
			test	0.9955	2.3021	17.6914	2.3022
			total	0.9976	1.7079	14.5621	1.7073