Forecast reconciliation: a review

George Athanasopoulos; Rob J. Hyndman; Nikolaos Kourentzes; Anastasios Panagiotelis

doi:10.1016/j.ijforecast.2023.10.010

Forecast reconciliation: a review

Econometrics & Business Statistics

Research output: Contribution to journal › Article › Research › peer-review

53 Citations (Scopus)

Abstract

Collections of time series formed via aggregation are prevalent in many fields. These are commonly referred to as hierarchical time series and may be constructed cross-sectionally across different variables, temporally by aggregating a single series at different frequencies, or even generalised beyond aggregation as time series that respect linear constraints. When forecasting such time series, a desirable condition is for forecasts to be coherent: to respect the constraints. The past decades have seen substantial growth in this field with the development of reconciliation methods that ensure coherent forecasts and improve forecast accuracy. This paper serves as a comprehensive review of forecast reconciliation and an entry point for researchers and practitioners dealing with hierarchical time series. The scope of the article includes perspectives on forecast reconciliation from machine learning, Bayesian statistics and probabilistic forecasting, as well as applications in economics, energy, tourism, retail demand and demography.

Original language	English
Pages (from-to)	430-456
Number of pages	27
Journal	International Journal of Forecasting
Volume	40
Issue number	2
DOIs	https://doi.org/10.1016/j.ijforecast.2023.10.010
Publication status	Published - Apr 2024

Keywords

Aggregation
Coherence
Cross-temporal
Grouped time series
Hierarchical time series
Temporal aggregation

Access to Document

10.1016/j.ijforecast.2023.10.010Licence: CC BY

565992428-oaFinal published version, 696 KBLicence: CC BY

Cite this

@article{629b4223052247078cd1084de6f75506,

title = "Forecast reconciliation: a review",

abstract = "Collections of time series formed via aggregation are prevalent in many fields. These are commonly referred to as hierarchical time series and may be constructed cross-sectionally across different variables, temporally by aggregating a single series at different frequencies, or even generalised beyond aggregation as time series that respect linear constraints. When forecasting such time series, a desirable condition is for forecasts to be coherent: to respect the constraints. The past decades have seen substantial growth in this field with the development of reconciliation methods that ensure coherent forecasts and improve forecast accuracy. This paper serves as a comprehensive review of forecast reconciliation and an entry point for researchers and practitioners dealing with hierarchical time series. The scope of the article includes perspectives on forecast reconciliation from machine learning, Bayesian statistics and probabilistic forecasting, as well as applications in economics, energy, tourism, retail demand and demography.",

keywords = "Aggregation, Coherence, Cross-temporal, Grouped time series, Hierarchical time series, Temporal aggregation",

author = "George Athanasopoulos and Hyndman, \{Rob J.\} and Nikolaos Kourentzes and Anastasios Panagiotelis",

note = "Funding Information: We thank Tommaso Di Fonzo, Xiaoqian Wang and Daniele Girolimetto for providing helpful comments on an earlier draft of this paper. Rob J Hyndman was funded by the Australian Government through the Australian Research Council Industrial Transformation Training Centre in Optimisation Technologies, Integrated Methodologies, and Applications (OPTIMA) , Project ID IC200100009 . Publisher Copyright: {\textcopyright} 2023 The Author(s)",

year = "2024",

month = apr,

doi = "10.1016/j.ijforecast.2023.10.010",

language = "English",

volume = "40",

pages = "430--456",

journal = "International Journal of Forecasting",

issn = "0169-2070",

publisher = "Elsevier",

number = "2",

}

TY - JOUR

T1 - Forecast reconciliation

T2 - a review

AU - Athanasopoulos, George

AU - Hyndman, Rob J.

AU - Kourentzes, Nikolaos

AU - Panagiotelis, Anastasios

N1 - Funding Information: We thank Tommaso Di Fonzo, Xiaoqian Wang and Daniele Girolimetto for providing helpful comments on an earlier draft of this paper. Rob J Hyndman was funded by the Australian Government through the Australian Research Council Industrial Transformation Training Centre in Optimisation Technologies, Integrated Methodologies, and Applications (OPTIMA) , Project ID IC200100009 . Publisher Copyright: © 2023 The Author(s)

PY - 2024/4

Y1 - 2024/4

N2 - Collections of time series formed via aggregation are prevalent in many fields. These are commonly referred to as hierarchical time series and may be constructed cross-sectionally across different variables, temporally by aggregating a single series at different frequencies, or even generalised beyond aggregation as time series that respect linear constraints. When forecasting such time series, a desirable condition is for forecasts to be coherent: to respect the constraints. The past decades have seen substantial growth in this field with the development of reconciliation methods that ensure coherent forecasts and improve forecast accuracy. This paper serves as a comprehensive review of forecast reconciliation and an entry point for researchers and practitioners dealing with hierarchical time series. The scope of the article includes perspectives on forecast reconciliation from machine learning, Bayesian statistics and probabilistic forecasting, as well as applications in economics, energy, tourism, retail demand and demography.

AB - Collections of time series formed via aggregation are prevalent in many fields. These are commonly referred to as hierarchical time series and may be constructed cross-sectionally across different variables, temporally by aggregating a single series at different frequencies, or even generalised beyond aggregation as time series that respect linear constraints. When forecasting such time series, a desirable condition is for forecasts to be coherent: to respect the constraints. The past decades have seen substantial growth in this field with the development of reconciliation methods that ensure coherent forecasts and improve forecast accuracy. This paper serves as a comprehensive review of forecast reconciliation and an entry point for researchers and practitioners dealing with hierarchical time series. The scope of the article includes perspectives on forecast reconciliation from machine learning, Bayesian statistics and probabilistic forecasting, as well as applications in economics, energy, tourism, retail demand and demography.

KW - Aggregation

KW - Coherence

KW - Cross-temporal

KW - Grouped time series

KW - Hierarchical time series

KW - Temporal aggregation

UR - https://www.scopus.com/pages/publications/85181065888

U2 - 10.1016/j.ijforecast.2023.10.010

DO - 10.1016/j.ijforecast.2023.10.010

M3 - Article

AN - SCOPUS:85181065888

SN - 0169-2070

VL - 40

SP - 430

EP - 456

JO - International Journal of Forecasting

JF - International Journal of Forecasting

IS - 2

ER -

Forecast reconciliation: a review

Abstract

Keywords

Access to Document

Other files and links

ARC Training Centre in Optimisation Technologies, Integrated Methodologies, and Applications (OPTIMA)

Cite this

Forecast reconciliation: a review

Abstract

Keywords

Access to Document

Other files and links

Projects

ARC Training Centre in Optimisation Technologies, Integrated Methodologies, and Applications (OPTIMA)

Cite this