Fig. 12From: A novel approach to workload prediction using attention-based LSTM encoder-decoder network in cloud environmentWorkload prediction results with/without the scroll prediction method. a Origin model for 36-step prediction. origin-36: The red line in sub-figure (a). The origin-36 line shows the predicted values of the origin model for 36-step prediction not using scroll prediction. scroll-24&12: The red line in sub-figure. b Scroll on two sequences with 24-step and 12-step prediction, respectively. b The scroll-24&12 line shows the pre-dicted values obtained via scroll prediction by splitting the prediction sequence with a length of 12 into two sequences with lengths of 24 and 12. c Scroll on three sequences with 12-step prediction. ground truth: The black line in all three sub figures. Ground truth is the real work-load in the dataset. Scroll-12&12&12: The red line in sub-figure (c). The scroll-12&12&12 line shows the predicted values obtained via scroll prediction by splitting the prediction se-quence with a length of 12 into three sequences with lengths of 12Back to article page