HierarchicalForecast: A Reference Framework for Hierarchical Forecasting in Python
Kin G. Olivares, Azul Garza, David Luo, Cristian Challú, Max Mergenthaler, Souhaib Ben Taieb, Shanika L. Wickramasuriya, Artur Dubrawski
TL;DR
This paper presents HierarchicalForecast, a Python open-source reference framework for hierarchical time series forecasting that enforces coherence across aggregation levels. It combines processed public datasets, evaluation metrics, and a curated set of fast statistical baselines (e.g., AutoARIMA/AutoETS via statsforecast) with a suite of reconciliation methods (BottomUp, TopDown, MiddleOut, MinTrace, ERM) and probabilistic approaches (PERMBU, NORMALITY, BOOTSTRAP). The framework addresses Python-specific barriers to accessing econometric baselines and computational bottlenecks by providing minimal dependencies and a just-in-time-compiled core. An end-to-end usage example demonstrates forecasting eight months for 57 Labour-series with multiple reconciliation schemes and bootstrap intervals, illustrating practical benchmarking. The work aims to bridge statistical/econometric methods and ML forecasting, offering a common benchmark to spur reproducibility and collaboration in hierarchical forecasting research.
Abstract
Large collections of time series data are commonly organized into structures with different levels of aggregation; examples include product and geographical groupings. It is often important to ensure that the forecasts are coherent so that the predicted values at disaggregate levels add up to the aggregate forecast. The growing interest of the Machine Learning community in hierarchical forecasting systems indicates that we are in a propitious moment to ensure that scientific endeavors are grounded on sound baselines. For this reason, we put forward the HierarchicalForecast library, which contains preprocessed publicly available datasets, evaluation metrics, and a compiled set of statistical baseline models. Our Python-based reference framework aims to bridge the gap between statistical and econometric modeling, and Machine Learning forecasting research. Code and documentation are available in https://github.com/Nixtla/hierarchicalforecast.