DOCSEARCHbyAlgolia
Submit Docs

FoRecoPy: Forecast Reconciliation in Python

website·Indexed 30 days ago

https://danigiro.github.io/FoReco/index.html

A Python framework for forecast reconciliation providing tools for optimal least-squares forecast combination within cross-sectional and temporal frameworks. It includes modules for approximating forecast error covariance matrices, shrinkage estimators, and utility classes for managing aggregation relationships and linear constraints using JAX arrays.

Tokens
91.9K
Snippets
341
Records
603
Agent score
72%

What's inside FoRecoPy

  1. Overview of FoRecoPy for Forecast Reconciliation

    FoRecoPy is a Python library used for forecast reconciliation, a post-forecasting process that improves accuracy and aligns forecasts for systems of linearly constrained time series (such as hierarchical or grouped time series). It provides Python functionality similar to the R package FoReco.

    Currently supported features include:

    • Regression-based reconciliation (e.g., minimum trace).
    • Cross-sectional reconciliation for hierarchical, grouped, and linearly constrained time series.
    • Temporal reconciliation for multiple aggregation frequencies.

  2. Overview of FoReco forecast reconciliation

    FoReco is an R package for forecast reconciliation, a post-forecasting process used to improve accuracy and align forecasts for systems of linearly constrained (e.g., hierarchical or grouped) time series. It supports three primary frameworks: Cross-sectional, Temporal, and Cross-Temporal. Reconciliation methods are categorized into Classical (Top-down, Bottom-up, Middle-out), Regression-based (Least squares, LCC), and Probabilistic (MVN, Sample-based).

  3. Overview of nlcReco for Non-linearly Constrained Forecast Reconciliation

    nlcReco is an R package implementing optimal combination forecast reconciliation procedures (Girolimetto et al. 2025) designed for multiple time series. It allows for the application of both linear and non-linear constraints within a cross-sectional framework to ensure reconciled forecasts.

  4. Overview of FoReco forecast reconciliation

    FoReco (Forecast Reconciliation) is an R package used for post-forecasting to improve the accuracy of base forecasts for systems of linearly constrained time series (such as hierarchical or grouped series). It supports classical (bottom-up, top-down) and modern (optimal and heuristic combination) reconciliation procedures for cross-sectional, temporal, and cross-temporal constraints.

  5. Overview of FoReco forecast reconciliation functions

    The FoReco R package provides tools for forecast reconciliation to improve the accuracy of base forecasts in systems of linearly constrained (e.g., hierarchical or grouped) time series. It supports classical methods (bottom-up, top-down) and modern methods (optimal and heuristic combinations).

  6. Overview of the FoCo2 R package

    The FoCo2 package provides methods and tools for coherent forecast combination of multiple time series subject to linear constraints. It allows users to improve forecast accuracy by leveraging inputs from multiple forecasting models or experts while ensuring that the resulting forecasts satisfy specified linear aggregation relationships. Additionally, the package supports the imposition of linear inequality constraints, such as ensuring non-negativity of the forecasts.

  7. Overview of FoRecoML for Nonlinear Forecast Reconciliation

    FoRecoML is an R package designed for nonlinear forecast reconciliation using machine learning. It supports three primary frameworks: cross-sectional (based on Spiliotis et al. 2021), temporal, and cross-temporal (based on Rombouts et al. 2024).

  8. Overview of FoReco package for forecast reconciliation

    The FoReco (Forecast Reconciliation) package is used for post-forecasting reconciliation to improve the quality of base forecasts for systems of linearly constrained time series (such as hierarchical or grouped series). It supports cross-sectional, temporal, and cross-temporal reconciliation using classical (bottom-up, top-down), optimal, and heuristic combination procedures.

  9. Overview of FoCo² (Coherent Forecast Combination)

    FoCo² is an R forecasting package used to combine multiple time series forecasts from different experts while adhering to linear constraints. It employs a multi-task approach to produce coherent forecasts using either optimal or heuristic methods. The package supports both balanced inputs (where all experts forecast all variables) and unbalanced inputs (where experts forecast different subsets of variables). Additionally, it provides tools to estimate forecast error covariance matrices, supporting simple, (block) diagonal, and full covariance structures.

  10. Overview of FoRecoPy for forecast reconciliation

    FoRecoPy is a Python package designed for the forecast reconciliation of general linearly constrained multiple time series. It adjusts base forecasts to satisfy known linear relationships (such as hierarchical sums, economic account balances, or temporal aggregations) while minimizing the least-squares distance from the original forecasts. The library supports both projection and structural approaches, provides multiple solvers for scalability, and can enforce non-negativity on the resulting reconciled forecasts.

  11. Overview of FoRecoPy for Forecast Reconciliation

    FoRecoPy is a Python package used for forecast reconciliation, a post-forecasting process that improves accuracy and aligns forecasts for systems of linearly constrained time series (such as hierarchical or grouped time series). It provides Python functionality similar to the R package 'FoReco'.

    Currently supported capabilities include:

    • Regression-based reconciliation (e.g., minimum trace).
    • Cross-sectional reconciliation for hierarchical, grouped, and linearly constrained time series.
    • Temporal reconciliation for multiple aggregation frequencies.

  12. Overview of FoRecoML for nonlinear forecast reconciliation

    FoRecoML is an R package designed for nonlinear forecast reconciliation in systems of linearly constrained time series (such as hierarchical or grouped series). It improves forecast accuracy by aligning forecasts using Machine Learning (ML) across cross-sectional, temporal, and cross-temporal settings. It inherits time series processing capabilities from the FoReco package.