/mahts

Model Agnostic Hierarchical Time Series in Python

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

mahts

Build Status PyPI version PyPI download

Model Agnostic Hierarchical Time Series in Python

This package implements algorithms for Hierarchical Times Series reconciliation. It is based on the hts R package (Paper) but has the benefit of being agnostic to the forecasting model.

Key Features

  • Easy to use interface.
  • Standard top-down, bottom-up & forecast-proportion approaches.
  • Fast computation of optimal combination through sparse matrices.
  • Optimal combination approach supports WLS with custom weights.
  • Optimal combination approach supports bounded solutions through the Trust Region Reflective algorithm.
  • Agnostic to the forecasting model.

Installation

Via PyPI:

pip install mahts

Or you can clone this repository and install it from source:

python setup.py install