. 2015 Oct 8;13:68. doi: 10.1186/s40201-015-0227-6

Table 2.

Summary results for MLP-NN for simulating EC

		Transfer function			Optimized number of neurons		Performance							Errors (μmhos/cm)
Station	Total number of data	First hidden layer	Second hidden layer	Output layer	First hidden layer	Second hidden layer	Train		Validation		Test		MAE	Max absolute error prediction	Min absolute error prediction
Station	Total number of data	First hidden layer	Second hidden layer	Output layer	First hidden layer	Second hidden layer	RMSE	R²	RMSE	R²	RMSE	R²	MAE	Max absolute error prediction	Min absolute error prediction
Ghaleh-shahrokh	141	Tan-sigmoid	Log-sigmoid	Purelin	10	4	22.08	0.82	22.15	0.82	28.66	0.75	22.83	65.04	1.23
Regulator dam	145				17	8	16.06	0.62	16.14	0.57	14.43	0.67	11.13	64.29	1.18
Zamankhan-bridge	129				5	4	17.50	0.68	19.02	0.60	20.40	0.69	16.97	58.55	0.60
Kaleh-bridge	106				17	10	30.26	0.91	33.77	0.80	60.12	0.79	47.79	96.31	1.27
Lenj	110				13	13	98.20	0.90	116.33	0.90	242.89	0.83	202.82	420.00	6.82
Mousian	105				7	5	88.07	0.89	93.24	0.88	117.26	0.85	84.15	327.80	7.87
Chum-bridge	105				8	8	103.38	0.88	109.98	0.82	145.46	0.81	110.70	281.10	0.13
Varzaneh	131				8	9	2541.63	0.89	3242.06	0.87	3893.45	0.83	3046.60	9665.00	506.40