Neural Networks in the NBA

Overview

This repository holds my final year Data Science Individual Project (dissertation). I experiment with Neural Networks and Multivariate Logistical Regression to assess if ML is capable of being used within the NBA draft

Notebook Contents

• Introduction to the raw data
• Extensive data cleaning/manipulation
• Dimensionality Reduction - PCA
• Models: Multi-Layered Perceptron (Neural Network) & Multivariate Logistical Regression
• Model Results & Interpretation: Standard evaluation metrics used (f1-score, recall, precision, accuracy) and shapley values used to bring deeper insight to the models

Final Report

• Introduction
  • 1.1 The NBA Draft and its importance ... 2
  • 1.2 Machine Learning ... 3
• Literature review overview and aim of project ... 4
  • 2.1 How ML is used in team selections in sports ... 4
    • 2.1.1 Neural Network Regressions ... 4
    • 2.1.2 Neural Network Classifications ... 5
  • 2.2 How NBA teams draft players ... 5
  • 2.3 Project steps and objectives ... 7
• Experimental Design ... 7
  • 3.1 Technology choice ... 7
  • 3.2 The Data ... 8
  • 3.3 Data preprocessing ... 8
    • 3.3.1 The Final Dataset ... 9
  • 3.4 Dimensionality Reduction ... 9
    • 3.4.1 Principal Component Analysis ... 10
    • 3.4.2 Shapley Values ... 11
  • 3.5 Neural Network - MLP ... 12
  • 3.6 Multivariate-Logistical Regression Classifier ... 13
  • 3.7 Evaluation metrics ... 13
• Results and Analysis ... 14
  • 4.1 My Results ... 14
  • 4.2 Support Vector Machines ... 17
  • 4.3 Regression Vs Classification ... 17
• Conclusion ... 18
  • 5.1 Aim of the project and intermediate steps ... 18
  • 5.2 Further steps ... 18
  • 5.3 Final thoughts ... 19

How to Use

1. Ensure you have Jupyter Notebook or JupyterLab installed on your machine (or run on Google Colab).
2. Clone/download this repository to your local machine.
3. Navigate to the directory containing this notebook.
4. Open the notebook via Jupyter interface.

Dependencies

• Python 3.10
• Libraries: numpy, pandas, matplotlib, shapley,sk-learn (ensure these are installed using pip install).

Contributing

Contributions to this project are welcome! Please fork the repository, make your changes, and submit a pull request.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Authors

• Joshua Adebayo BSc Data Science @ University of Exeter 2024

Acknowledgments

• Thanks to my supervisor David Walker