A Multivariate and Multistage Medium- and Long-Term Streamflow Prediction Based on an Ensemble of Signal Decomposition Techniques with a Deep Learning Network

Table 3

Comparison of models using statistical metrics during the testing period.

Model	MAE (m^3/s)	RMSE (m^3/s)	NSCE	MAPE (%)
RBF	41.577	64.280	0.829	34.749
SVR	39.010	61.075	0.846	33.035
RFR	36.069	54.871	0.876	31.021
GRU	33.800	50.485	0.895	29.453
LSTM	28.745	42.674	0.925	27.822
ICEEMDAN-LSTM	24.613	32.418	0.957	26.262
VMD-LSTM	20.569	27.642	0.968	24.113
IVL (ICEEMDAN-VMD-LSTM)	10.075	19.249	0.985	9.050