We are inviting submissions to a special issue at the International Journal of Forecasting on the topic of “Innovations in Hierarchical Forecasting”. The special issue is guest edited by G. Athanasopoulos, R. J. Hyndman, A. Panagiotelis and myself and its submission deadline is on the 31st of August 2021.
Organisations make multiple decisions informed by forecasts both to plan and function efficiently. Such decisions may differ in scope, from operational, to tactical, to strategic; corresponding to different time scales from short-term to medium-term to long-term; and can have different foci, for example inventory control for a single product, for a single store, or across an entire supply chain. Organisations that face such challenges include businesses, not-for-profit organisations, and policymakers who address societal challenges.
Forecasts of quantities that adhere to some known constraints should be coherent; that is, the predicted values at disaggregate scales should add up to the aggregate forecast. For example, monthly predictions should sum up to annual predictions and similarly, regional predictions should add up to country-level predictions. This is an important qualifier for forecasts to support aligned decision making across different planning units and horizons.
This forecasting problem setting gives rise to hierarchical forecasting. Historically this has been addressed using Top-Down and Bottom-Up approaches, which have been shown to exhibit several limitations. In the past decade, the introduction of forecast reconciliation approaches has reinvigorated research into hierarchical forecasting. Recent work has looked at novel estimation techniques, expansion of the hierarchical framework to temporal and cross-temporal hierarchies, probabilistic hierarchical forecasting, alternative understandings of the problem through a geometric or optimisation lens, amongst many other contributions. Meanwhile, forecast reconciliation techniques have been applied to a number of novel domains including energy, macroeconomics, mortality, tourism, and intermittent demand.
Areas of interest include, but are not limited to:
- New methodologies for hierarchical forecasting
- High dimensional hierarchical forecasting (methods and applications)
- An improved understanding of the relationship between forecast reconciliation and forecast combination
- Probabilistic hierarchical forecasting
- Temporal and cross-temporal hierarchies
- Machine learning and AI approaches to hierarchical forecasting
- Hierarchical forecasting with explanatory variables
- Applications of hierarchical forecasting to new domains
We would like to see submissions from diverse backgrounds, reflecting the nature of forecasting in organisations. You can find additional information here.