Combining LSTM Network Ensemble via Adaptive Weighting for Improved Time Series Forecasting

Table 3

Comparison with other state-of-the-art ensemble forecasting methods. (a) MAE and (b) MSE.

(a)

Time series	Ensemble forecasting methods
	Avg.[3–5]	Median[15]	LSR[2]	AIW[16]	NNLE[6]	Proposed
River Flow	0.751	0.676	0.736	0.749	0.638	0.536
Vehicles	2.087	2.139	2.059	2.071	2.001	1.851
Wine	2.075	2.173	2.466	2.372	1.923	0.937
Airline	11.63	11.73	10.68	10.22	7.434	5.582

(b)

Time series	Ensemble forecasting methods
	Avg.[3–5]	Median[15]	LSR[2]	AIW[16]	NNLE[6]	Proposed
River Flow	1.158	1.138	1.245	1.156	0.978	0.881
Vehicles	6.188	6.683	7.508	6.175	5.531	4.452
Wine	9.233	10.05	10.07	9.204	7.524	3.653
Airline	181.4	176.5	158.3	143.1	86.63	64.112

Note: Bold values denote the best results of forecasting each time series.