Improving forecasting by estimating time series structural components across multiple frequencies

A novel algorithm is proposed that aims to mitigate the importance of model selection, while increasing the accuracy of forecasting accuracy, by constructing multiple time series from the original time series, using temporal aggregation.