Fraud detection in financial transactions using explainable ML & DL models

This repository contains the final project for the DSAI 305 course at Zewail City, focusing on explainable AI models for financial transaction fraud detection. The final_models/ directory includes Jupyter notebooks, each implementing a machine learning or deep learning model with explainability techniques.

Instructions

1. Open Notebooks:

   • Navigate to the final_models/ directory.
   • Open each Jupyter notebook and preferably, use GPU powered environment in a Jupyter environment (e.g., JupyterLab, Google Colab).

2. Run All Cells:

   • Execute cells sequentially, as data, imports and dependencies are defined at the top.
   • Each notebook covers data preprocessing, model training, evaluation, and explainability, so review the Table of Contents.

Notes

• Ensure dependencies (pandas, numpy, scikit-learn, imblearn, lightgbm, tensorflow, shap, lime) are installed via pip.
• Datasets are loaded within notebooks (e.g., PaySim via kagglehub). (Imported automatically in each notebook).
• For resource-intensive explainability sections (e.g., SHAP, LIME), adjust sample sizes if needed.