Football Game Result Prediction

This repository provides a solution for predicting football game results. The package leverages various tools and frameworks including Scikit-learn, MLflow, FastAPI, Streamlit, and Docker. It encompasses the entire machine learning lifecycle, from data loading and preprocessing to model training, saving, and deployment.

Steps of the solution

• Load data from a PostgreSQL database
• Preprocess data using Pandas and Scikit-learn
• Train a model using XGBoost
• Save the trained model to MLflow registry
• Serve model inference using FastAPI
• Interface the FastAPI application with a Streamlit front-end
• Dockerized environment for seamless deployment
• Installation

Clone the repository

git clone https://github.com/yourusername/game-prediction.git
cd game-prediction

Set up a virtual environment and install dependencies

Before running any script, you will need your own version of the required database, using another repository : https://github.com/Sonson6/football-analytics.git

Once you have it, you must create a .env file in the repository and fill the following variables :

DATABASE_USER=your_postgres_username
DATABASE_PASSWORD=your_postgres_password
DATABASE_HOST=host.docker.internal # no need to touch it if being run locally
DATABASE_NAME=your_database_name

Now, you can install the dependencies :

poetry install

Train your own model :

poetry run game_prediction/train.py

Finally, you can create the Docker image and run the containers :

docker compose up --build

Now, you can play with the Streamlit interface to get model predictions (will be accessible through http://localhost:8501).

Contributing

Contributions are welcome! Please submit a pull request or open an issue to discuss improvements or features.