Hierarchical forecast reconciliation with machine learning

Spiliotis, Evangelos; Abolghasemi, Mahdi; Hyndman, Rob J.; Petropoulos, Fotios; Assimakopoulos, Vassilios

doi:10.1016/j.asoc.2021.107756

Cited by 42 publications

(27 citation statements)

References 44 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Empirical results in Wickramasuriya et al (2019) and Panagiotelis et al (2021) show that reconciled forecasts, overall, tend to be, on average, more accurate than the original (base) forecasts. Spiliotis et al (2021) introduce machine learning models to the reconciliation procedure and allow non-linear combinations of the original (base) forecasts. 1…”

Section: Forecast Reconciliationmentioning

confidence: 99%

“…Briefly, their argument is that series with smaller errors are expected to have more accurate forecasts, so they should have larger weights, and they will be changed little during reconciliation. On the other hand, Jeon et al ( 2019) and Spiliotis et al (2021) recommend using the cross-validation procedure and the resulting out-of-sample errors in obtaining the weights. They note that insample errors do not reliably proxy out-of-sample errors.…”

Section: Figure 3 Is Herementioning

confidence: 99%

See 1 more Smart Citation

Counterfactual reconciliation: Incorporating aggregation constraints for more accurate causal effect estimates

Cengiz

Tekgüç

2024

International Journal of Forecasting

View full text Add to dashboard Cite

Section: Forecast Reconciliationmentioning

confidence: 99%

Section: Figure 3 Is Herementioning

confidence: 99%

Counterfactual reconciliation: Incorporating aggregation constraints for more accurate causal effect estimates

Cengiz

Tekgüç

2024

International Journal of Forecasting

View full text Add to dashboard Cite

“…Generating base forecasts for each series implies that specialized models can be developed for each part of the hierarchy, incorporating node-specific available information [2]. Base forecasts are then linearly combined (reconciled) leveraging available information across the hierarchy to ensure coherency; a process employed by all hierarchical forecasting approaches as of to date [1,3,5,6,11,12,13,14,15,16,17,18,19,20,21].…”

Section: Hierarchical Forecastingmentioning

confidence: 99%

“…Relying on powerful statistical regressors and the availability of larger and richer data sets, machine learning emerges as an appealing and suitable tool for estimating the persistently challenging covariance matrix. Spiliotis et al [5] put forward such an approach employing a bottom-up method to reconcile predictions from Random Forest and XGBoost regressors. Taking as input the base forecasts of all the series of the hierarchy, the reconciled tree is then obtained from bottom-up aggregation.…”

Section: Reconciliation Approachesmentioning

confidence: 99%

“…When dealing with different abstraction levels across a system or organization, it is commonly more important to obtain coherent predictions across all considered layers and horizons, not to result in unaligned decisions or possibly even conflicting ones [1]. This obstacle arises in multiple domains, including tourism [2,3], retail [4], stock management [5] and smart grid management [6], which showcases this matter quite adequately.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Hierarchical learning, forecasting coherent spatio-temporal individual and aggregated building loads

Leprince¹,

Madsen²,

Møller³

et al. 2023

Preprint

View full text Add to dashboard Cite

Optimal decision-making compels us to anticipate the future at different horizons. However, in many domains connecting together predictions from multiple time horizons and abstractions levels across their organization becomes all the more important, else decision-makers would be planning using separate and possibly conflicting views of the future. This notably applies to smart grid operation. To optimally manage energy flows in such systems, accurate and coherent predictions must be made across varying aggregation levels and horizons. Such hierarchical structures are said to be coherent when values at different scales are equal when brought to the same level, else would need to be reconciled. With this work, we propose a novel multi-dimensional hierarchical forecasting method built upon structurally-informed machine-learning regressors and established hierarchical reconciliation taxonomy. A generic formulation of multi-dimensional hierarchies, reconciling spatial and temporal hierarchies under a common frame is initially defined. Next, a coherency-informed hierarchical learner is developed built upon a custom loss function leveraging optimal reconciliation methods. Coherency of the produced hierarchical forecasts is then secured using similar reconciliation technics. The outcome is a unified and coherent forecast across all examined dimensions, granting decision-makers a common view of the future serving aligned decision-making. The method is evaluated on two different case studies to predict building electrical loads across spatial, temporal, and spatio-temporal hierarchies, benchmarked against base and multi-task regressors. Although the regressor natively profits from computationally efficient learning thanks to the unification of independent forecasts into a global multi-task model, results displayed disparate performances, demonstrating the value of hierarchical-coherent learning in only one setting. Yet, supported by a comprehensive result analysis, existing obstacles were clearly delineated, presenting distinct pathways for future work. Particularly, investigating reduced number of model weights and varying native multi-output regressors can tackle the two preeminent challenges that are the curse of dimensionality and coherency-learning complications from scaled trees respectively. Overall, the paper expands and unites traditionally disjointed hierarchical forecasting methods providing a fertile route toward a novel generation of forecasting regressors.

show abstract

Commentary on multivariate dynamic modeling for Bayesian forecasting of business revenue

Yelland

2022

Appl Stoch Models Bus & Ind

View full text Add to dashboard Cite

Hierarchical forecast reconciliation with machine learning

Cited by 42 publications

References 44 publications

Counterfactual reconciliation: Incorporating aggregation constraints for more accurate causal effect estimates

Counterfactual reconciliation: Incorporating aggregation constraints for more accurate causal effect estimates

Hierarchical learning, forecasting coherent spatio-temporal individual and aggregated building loads

Commentary on multivariate dynamic modeling for Bayesian forecasting of business revenue

Contact Info

Product

Resources

About