Environmental Sciences Europe

Table 1 Performance of the 7-day ahead forecasts using the LSTM, MLP, Linear and Baseline models on the test set over different forecasting horizons (1‒7 days) using different evaluation metrics: MAE, RMSE, R² correlation, and WI

From: Water level prediction using long short-term memory neural network model for a lowland river: a case study on the Tisza River, Central Europe

Day	MAE (cm)				RMSE (cm)
Day	LSTM	MLP	Linear	Baseline	LSTM	MLP	Linear	Baseline
First	4.2	9.9	7.7	9.7	6.0	13.7	10.3	15.2
Second	7.6	12.3	11.4	18.4	11.4	16.6	15.5	29.0
Third	11.3	14.9	16.0	26.4	17.6	20.6	22.3	41.4
Fourth	16.3	19.6	21.6	33.5	26.2	28.1	30.5	52.4
Fifth	22.4	25.6	28.6	40.0	36.3	37.5	40.2	62.2
Sixth	28.7	32.2	36.1	45.8	46.3	47.1	50.5	70.9
Seventh	34.7	38.4	41.4	51.1	55.5	55.8	58.4	78.7

Day	R² (−)				WI (−)
Day	LSTM	MLP	Linear	Baseline	LSTM	MLP	Linear	Baseline
First	0.9987	0.9935	0.9963	0.9920	0.9988	0.9937	0.9964	0.9920
Second	0.9955	0.9906	0.9918	0.9709	0.9956	0.9907	0.9919	0.9711
Third	0.9893	0.9854	0.9829	0.9408	0.9894	0.9854	0.9828	0.9410
Fourth	0.9763	0.9728	0.9680	0.9051	0.9765	0.9725	0.9676	0.9056
Fifth	0.9545	0.9516	0.9443	0.8666	0.9543	0.9496	0.9419	0.8672
Sixth	0.9260	0.9233	0.9122	0.8268	0.9243	0.9172	0.9061	0.8276
Seventh	0.8937	0.8925	0.8824	0.7865	0.8889	0.8790	0.8693	0.7874

Back to article page