This repository contains the implementation of a Graph Attention Network (GAT) designed as part of an Undergraduate Research Scholar (URS) thesis at Texas A&M University. The primary aim of this project is to assess the efficacy of the GAT model in predicting illicit transactions on the Bitcoin blockchain in comparison to other binary classification models.

The data utilized for this project originates from the Elliptic dataset. This dataset provides labeled transactional data for illicit and licit Bitcoin transactions, serving as an ideal benchmark for our model.

• TensorFlow
• Keras
• NumPy
• Pandas
• scikit-learn

1. Ensure all dependencies are installed.
2. Download the Elliptic dataset and place the required CSV files in the project directory.
3. Run gat.py to train and evaluate the GAT model.

This project was conducted as part of an Undergraduate Research Scholar (URS) thesis at Texas A&M University. A special thanks to all mentors and peers who provided valuable insights and feedback throughout the research journey.