Probabilistic reconciliation of time series forecasts

Reconcile implements probabilistic time series forecast reconciliation methods introduced in

1. Zambon, Lorenzo, Dario Azzimonti, and Giorgio Corani. "Probabilistic reconciliation of forecasts via importance sampling." arXiv preprint arXiv:2210.02286 (2022).
2. Panagiotelis, Anastasios, et al. "Probabilistic forecast reconciliation: Properties, evaluation and score optimisation." European Journal of Operational Research (2022).

The package implements methods to compute summing/aggregation matrices for grouped and hierarchical time series and reconciliation methods for probabilistic forecasts based on sampling and optimization, and in the near future also some recent forecasting methods, such as proposed in Benavoli, et al. (2021) or Corani et al., (2020) via GPJax.

An example timeseries forecast application using GPs can be found in examples/reconciliation.py and a case study on probabilistic forecast reconciliation of stock index data can be found here.

Make sure to have a working JAX installation. Depending whether you want to use CPU/GPU/TPU, please follow these instructions.

To install the package from PyPI, call:

pip install probabilistic-reconciliation

To install the latest GitHub , just call the following on the command line:

pip install git+https://github.com/dirmeier/reconcile@<RELEASE>

Simon Dirmeier sfyrbnd @ pm me