A novel approach to workload prediction using attention-based LSTM encoder-decoder network in cloud environment

EURASIP Journal on Wireless Communications and Networking

Table 4 RMSE of different history window lengths for the training set and validation set when the prediction sequence length is 12

History window length	Training×10⁻²	Validation×10⁻²	Relative ratio over the best
12	2.165	4.308	4.21%
18	1.819	4.134	0
24	1.786	4.137	0.07%
30	1.792	4.135	0.02%
36	1.753	4.141	0.16%
42	1.757	4.164	0.73%
48	1.751	4.187	1.28%