A1638
Title: Hierarchies everywhere: Managing \& measuring uncertainty in hierarchical time series
Authors: Ross Hollyman - University of Exeter (United Kingdom) [presenting]
Abstract: The problem of making reconciled forecasts of large collections of related time series through a behavioural / Bayesian lens is examined. The approach explicitly acknowledges and exploits the "connectedness" of the series in terms of time-series characteristics, forecast accuracy, and the hierarchical structure. By maximising the available information and significantly reducing the dimensionality of the hierarchical forecasting problem, it is shown how to improve the accuracy of the reconciled forecasts. In contrast to existing approaches, the structure allows the analysis and assessment of the forecast value added at each hierarchical level. The reconciled forecasts are inherently probabilistic, whether probabilistic base forecasts are used or not.
