We developed and carried out a synthetic-information-generation system to further more Appraise the performance on the proposed model in the existence of different seasonal components.

A solitary linear layer is sufficiently robust to design and forecast time series knowledge furnished it has been appropriately decomposed. Therefore, we allotted one linear layer for every element in this examine.

We produce a time series with hourly frequency which has a day-to-day and weekly seasonality here which observe a sine wave. We exhibit a more real world example afterwards while in the notebook.

Home windows - The lengths of each seasonal smoother with respect to each period. If they are massive then the seasonal element will present fewer variability with time. Must be odd. If None a list of default values determined by experiments in the first paper [1] are utilized.