Forecasting gas demand distributed via local gas networks in the UK.
All analysis and results are contained in the notebook main.ipynb. The source directory src/ contains some helper code, including the Bayesian linear regression model, linear basis mapping and some evaluation metrics. The provided datasets can be found in the data/ directory.
A good reference on BLR can be found in chapter 9 of Maths for Machine Learning.
Apologies for the lack of docstrings and proper commenting / documentation, I was quite pressed for time!
To install the environment, you will need to have python 3.10 installed as well as poetry installation. The tensorflow dependencies are for MacOS -- simply replace tensorflow-macos with tensorflow in the pyproject.toml file for installation on Linux or Windows. From the top-level directory, run the following commands:
poetry env use <path to python 3.10 executable>
poetry installWith the environment activated, run:
jupyter notebook