Improved Rainfall Prediction through Nonlinear Autoregressive Network with Exogenous Variables: A Case Study in Andes High Mountain Region

Table 5

Results for NARX net models.

Basin	Model	Subset	NSE	KGE	PBIAS	RMSE
Labrado	3 lags, 50 hidden neurons	Train (Tr)	0.57	0.77	−0.50	31.20
		Cross-validation (Cv)	0.38	0.70	0.60	30.83
		Test (Ts)	0.54	0.77	−3.00	30.02
		Test close-loop	0.57	0.51	12.30	47.74
Chirimachay	6 lags, 50 hidden neurons	Train (Tr)	0.52	0.75	2.60	42.72
		Cross-validation (Cv)	0.53	0.72	−5.20	46.20
		Test (Ts)	0.54	0.69	5.10	36.69
		Test close-loop	0.61	0.61	22.20	36.76