Skip to content

README.md

Latest commit

 

History

History
113 lines (87 loc) · 2.71 KB

README.md

File metadata and controls

113 lines (87 loc) · 2.71 KB

⚽ Premier League Match Prediction

A full-stack machine learning web app that predicts English Premier League match outcomes — given two teams, it returns the most likely result (Win, Draw, or Lose) based on recent form and head-to-head history.

🛠️ Technologies Used

  • 🐍 Python 3.10+
  • 🔥 Flask — lightweight backend API
  • 🤖 scikit-learn, Random Forest, Gradient Boost, XGBoost — ML model training and prediction
  • 📊 pandas, NumPy — data preprocessing
  • 🎯 GridSearchCV — model tuning
  • 🎨 HTML + CSS + JavaScript — responsive, interactive UI
  • 🧠 joblib — model serialization
  • 🌐 CORS — cross-origin communication
  • 📈 Matplotlib / Seaborn — model evaluation (heatmaps, correlation)

✨ Features

  • 📊 Recent 5-game rolling averages (GF, GA, xG, SoT, FK, etc.)
  • 🔁 Head-to-head record encoding & draw rate calculations
  • 🔄 Bi-directional prediction logic for better balance
  • 📦 REST API returning predictions & probability distribution
  • 🎨 Clean, responsive team selector interface with logos & colors
  • 🧪 Graceful fallback for teams with no match history due to promotion/relegation

📦 Local Setup

1. Clone the Repo

git clone https://github.com/your-username/Premier-League-Prediction.git
cd Premier-League-Prediction

2. Create and Activate a Virtual Environment

python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate

3. Install Dependencies

pip install -r requirements.txt

🚀 Running the App

python app.py

App will be live at: http://localhost:5000

📡 API Usage

POST /predict

Request:

{
  "team1": "Manchester City",
  "team2": "Liverpool"
}

Response:

{
  "result": 2,
  "probabilities": [0.18, 0.25, 0.57]
}
  • result: 0 = Lose, 1 = Draw, 2 = Win (for team1)
  • probabilities: order is [Lose, Draw, Win]

🧠 Project Structure

premier-prediction/
├── app.py                 # Flask backend entry
├── premier_model.pkl      # Trained model
├── prem_data.csv          # Cleaned match data
├── requirements.txt
├── notebooks/             
│   ├── Prem_Prediction_Scraping.ipynb     # Web scraping logic
│   └── Prem_Prediction_Model.ipynb        # EDA, feature engineering, model prototyping
├── static/                
│   ├── script.js
│   ├── styles.css
│   └── images/            # team logo png files
├── templates/
│   └── index.html
├── src/
│   ├── preprocessor.py
│   ├── train_model.py
│   └── infer.py

📝 License

MIT License © 2025 Aiden Cho