Forecast reconciliation is a post-forecasting process aimed at improving the accuracy and coherence of forecasts for a system of linearly constrained time series (e.g., hierarchical, grouped, or temporal structures).
The FoRecoPy package is inspired by the R package FoReco and brings similar functionality to Python. It is designed for researchers, practitioners, and data scientists who use Python for time series forecasting and want access to state-of-the-art reconciliation methods.
Future versions will expand the scope to include the cross-temporal framework, non-negative constraints and probabilistic reconciliation.
Make sure to have a working JAX installation (please, follow these instructions).
To install the package from PyPI, call:
pip install forecopyTo install the latest GitHub , just call the following on the command line:
pip install git+https://github.com/danigiro/FoRecoPy@<RELEASE>- Optimal combination reconciliation via projection and structural approaches
- Tools for both cross-sectional (
csrec) and temporal (terec) reconciliation - Different covariance matrix approximation
- Support for custom aggregation or constraints matrices
- Option to enforce non-negativity on reconciled forecasts
- Efficient solvers suitable for high-dimensional problems
Examples of cross-sectional and temporal forecast reconciliation are available here
MIT License. See LICENSE for details.