football_ai

Football Match Predictor ⚽

This project is a machine learning-based web application for predicting the outcome of football matches. It uses data from FBref and is built with Streamlit to create an interactive and user-friendly interface. The model is trained using GradientBoostingClassifier, and the app allows users to select home and away teams to predict match outcomes.

Features

• Football Match Predictions: Predicts whether the home team will win, the away team will win, or if the match will end in a draw.
• Interactive UI: Built using Streamlit, with a responsive design and smooth team selection.
• Custom Styling: The UI features a custom CSS for an enhanced user experience, including gradient backgrounds and a clean layout.

Data Source

The match data used for training the model was sourced from FBref, which contains a rich set of football statistics for teams and players.

Installation

Prerequisites

Before running the app, ensure you have Python 3.11.7 installed.

Virtual Environment Setup

Create a virtual environment using Conda:

conda create -p venv python==3.11.7

Activate the environment:

conda activate ./venv

Installing Dependencies

Install the necessary packages listed in requirements.txt:

pip install -r requirements.txt

Alternatively, the main libraries used are:

• Streamlit: For creating the web app interface
• Pandas: For data manipulation
• NumPy: For numerical operations
• Scikit-learn: For machine learning models

You can install them individually if needed:

pip install streamlit pandas numpy scikit-learn

How to Run

To start the app, run the following command:

streamlit run app.py

This will launch the web app in your browser, where you can select home and away teams and predict the match outcome.

Model Information

The model used for prediction is a GradientBoostingClassifier. It has been trained on football data to predict whether the home team wins, the away team wins, or the match ends in a draw. The training data includes historical match statistics and results from various teams in the Premier League.

Usage

1. Select Teams: Choose the home and away teams from the dropdown menus.
2. Predict Outcome: Click the Predict Outcome button, and the model will predict the match result based on the selected teams.
3. Results: The predicted outcome will be displayed below the selection in a user-friendly format (e.g., Home Win, Away Win, or Draw).

Customization

• Team List: The teams available for selection are defined in the teams_dict dictionary. You can easily update or modify these teams as needed.
• Styling: Custom CSS is used to style the app. If you'd like to change the appearance, modify the CSS code in the Streamlit st.markdown block.

Project Structure

football-predictor/
│
├── main.py               # The main Streamlit app file
├── model.pkl            # The pre-trained machine learning model
├── database.csv         # The dataset containing match statistics
├── venv/                # Conda virtual environment
└── requirements.txt     # List of dependencies for the project

Acknowledgments

• Football match data sourced from FBref.
• Built using Streamlit, Pandas, and Scikit-learn.

Future Improvements

• Add more leagues and teams to the predictor.
• Improve the model by incorporating additional features such as player form and injuries.
• Enable real-time match updates and predictions.

---