Python code and Jupyter Notebooks for a pet project to predict football match results using various statistical models - both classical and ML
Code consists of several main branches:
- Database Creation/Maintenance: Code to create and regularly maintain an up to date sqlite database of European Football Matches pulled from here
- Feature Creation/Maintenance: Code to take this raw match result data and create relevant features for use in predictive modelling
- Modelling/Evaluation Helpers: Helper functions wrapped around standard libs (statsmodels, sklearn) to improve notebook workflow (e.g. pretty print confusion matrix using seaborn)
Dependencies
- git: Version Control System - used here to clone the git repo locally. Pre-installed on most linux/mac, install instructions here if required
- conda: Package & environment manager - used to create python environment (populated with required python libs) to run the code. Helpful installation instructions for miniconda here
Instructions
To run the code on your local machine from scratch:
- Once in your chosen parent directory, run the following to clone this git repo locally (requires git to be installed - comes as default on mac):
git clone https://github.com/mjam03/epl.git
- Create a conda environment to run the code inside - this is initialised by the included
environment.ymlfile:
conda env create -f environment.yml
- Activate the newly created environment - this is required before the next step so
eplis initialised within the conda env:
conda activate epl
- Inside directory (i.e. in
./epl/), execute the setup.py script s.t. the python functions withinepl/can be picked up - explained brilliantly here:
pip install -e .
pip does this for us and will create a .egg-info/ directory (used internally to identify the files needed), the -e arg ensures local changes to the code are picked up (do not need to pip install again to pick up changes)
- Create the sqlite database by running the
parse.pyscript insideepl/scripts/:
cd ./epl/scripts/
python parse.py