Using Hybrid Wavelet-Exponential Smoothing Approach for Streamflow Modeling

Table 6

Performance results of AI-based models.

	Watersheds	Model	Inputs	Best structures	No. of hidden neurons	Epoch	NSE		RMSE (m^3/s)		NSEpeak
							Training	Test	Training	Test	Test
Daily	Trinity River	ANN	Q (t), Q (t−1), Q (t−2)	3-4-1	4	20	0.852	0.845	39.938	72.981	0.791
		WANN	Qa (t), Qd2 (t), Qa (t−1), Qd4 (t−1)	4-7-1	7	30	0.932	0.921	30.471	51.305	0.905
	West Nishnabotna River	ANN	Q (t), Q (t−1), Q (t−2)	3-9-1	9	90	0.689	0.658	19.407	19.807	0.569
		WANN	Qa (t), Qd1 (t), Qa (t−1), Qd1 (t−1)	4-9-1	9	40	0.823	0.812	17.022	17.326	0.765
Monthly	Trinity River	ANN	Q (t), Q (t−12)	2-5-1	5	20	0.611	0.601	2123.189	2401.262	0.679
		WANN	Qa (t), Qd3 (t), Qa (t−1), Qd4 (t−1)	4-4-1	4	10	0.85	0.822	1381.341	1877.409	0.851
	West Nishnabotna River	ANN	Q (t), Q (t−12)	2-7-1	7	100	0.578	0.545	470.929	517.635	0.559
		WANN	Qa (t), Qd4 (t), Qa (t−1), Qd3 (t−1)	4-8-1	8	10	0.779	0.764	258.957	411.186	0.761