Skip to content

Latest commit

 

History

History
90 lines (64 loc) · 3.84 KB

Readme.md

File metadata and controls

90 lines (64 loc) · 3.84 KB

🚀 Transforming Fantasy Premier League (FPL) Data into Actionable Player Insights with Machine Learning 🚀

I'm thrilled to share a project that combines machine learning and data engineering to revolutionize Fantasy Premier League (FPL) transfers! This recommendation system leverages unsupervised learning techniques—specifically KMeans clustering and Principal Component Analysis (PCA)—to offer FPL managers deeper insights into player performance and transfer strategies.

🔍 Project Highlights:

  • Unsupervised Learning: Unlike traditional supervised approaches, this project uncovers hidden patterns and relationships in the data without predefined labels, offering fresh insights into player potential.
  • Variance-Weighted PCA Scoring: Using PCA, complex player metrics are distilled into principal components, ranked by variance-weighted scores for more effective player rankings.
  • Advanced Features: The model incorporates a variety of features, including:
    • Goals, assists, clean sheets, minutes played
    • Bonus points, influence, creativity, threat levels
    • ICT index, fixture difficulty rating (FDR)
    • Smart-engineered features like penalty and freekick orders
  • Interactive System: A Flask-powered app allows users to input their team ID and receive tailored transfer recommendations.

💡 Gameweek Updates: The tool balances historical points with fixture-adjusted potential, dynamically updating every gameweek to provide a strategic edge.

🛠️ Tech Stack

  • Python: Core programming language
  • KMeans Clustering: For player grouping
  • PCA: For dimensionality reduction and scoring
  • Flask: Backend framework for the interactive system
  • Matplotlib/Seaborn: For data visualization
  • SQLite: For storing player and team data

📸 Screenshots

1. Input Form

An interactive form where users enter their team ID to generate personalized recommendations: Input Form Screenshot

2. Player Recommendations

A detailed list of recommended players, ranked by variance-weighted PCA scores: Player Recommendations Screenshot 1 Player Recommendations Screenshot 2 Player Recommendations Screenshot 3

3. Expected Points Breakdown

A visualization of expected points for each recommended player, including metrics like goals, assists, and clean sheets: Expected Points Breakdown Screenshot

🚀 Getting Started

Prerequisites

  • Python 3.x installed
  • Libraries listed in requirements.txt

Installation

  1. Clone the repository: 
    git clone https://github.com/zobrathemanish/FPL-Recommendation-System.git
  2. Navigate to the Project directory 
    cd FPL-Recommendation-System
  3. Navigate to the Project directory 
    pip install -r requirements.txt
  4. Run the Flask Application 
    flask run

📈 Key Functionalities

  1. Player Clustering:
    Groups players into clusters based on performance metrics to identify hidden patterns.

  2. Variance-Weighted Ranking:
    Uses PCA to rank players based on their variance-weighted contribution to the principal components.

  3. Fixture-Adjusted Recommendations:
    Balances historical performance with fixture difficulty to suggest optimal transfers.

📢 Final Thoughts

This project is designed to empower FPL managers with a strategic edge, offering actionable insights that balance historical data and future potential. Whether you're aiming to dominate your mini-league or climb the global leaderboard, this recommendation system provides the tools you need to make data-driven decisions.

Check out the repository: FPL Recommendation System
Happy managing! ⚽