mahts

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