Inductive GMDH-Based Approach to Hierarchical Forecasting

Abstract

In the paper, the inductive algorithms of hierarchical modelling for long-term forecasting are considered. GMDH algorithms are used to get accurate long-term forecasts on all levels of temporal data. Inductive approach allows to reconcile models and to increase accuracy of forecasting simultaneously on all levels of modelling. The results show that multi-level inductive algorithms can improve quality and extend forecast horizon in comparison with conventional univariate methods used in the hierarchical forecasting.