Skip to content

Latest commit

 

History

History
104 lines (71 loc) · 3.35 KB

README.md

File metadata and controls

104 lines (71 loc) · 3.35 KB

BoostFex

🌟 Detection of Bank Transaction Anomalies Using Gradient Boosted Federated Learning 🌟

📚 Abstract

Official implementation of 'Detection of Bank Transaction Anomalies using Gradient Boosted Federated Learning' (IEEE Access). Explore cutting-edge privacy-preserving techniques, scalable federated learning models, and advanced anomaly detection algorithms for financial transactions.

🛠️ Overview

Fraudulent transactions pose critical challenges for the banking sector, affecting customer trust and financial security. In this research, we propose an innovative Federated Learning framework integrated with Gradient Boosting algorithms to enhance anomaly detection. Our approach ensures robust privacy-preserving mechanisms while delivering high accuracy on imbalanced datasets.

🔍 Key Features

  • Federated Learning Architecture: Leverages decentralized training to enhance data security.
  • Gradient Boosted Models: Optimized for imbalanced datasets and high-dimensional transaction data.
  • Privacy-Preserving Techniques: Compliance with GDPR and CCPA standards.
  • Scalable and Modular Design: Supports both cyclic and bagging federated learning models.

🏗️ Repository Structure

📁 src/                  # Implementation of different models
📁 Dataset/              # Dataset Generator and Preprocessing files
📁 Results/              # Raw result data and Graph Visualization
📁 Literature Survey     # Literature Survey Papers
📁 Journal Article       # Journal Article Submitted to IEEE Access
📄 README.md             # Project description

📈 Results Summary

Metric Cyclic Model Bagging Model Centralized Baseline
Accuracy 98% 97% 92%
Precision 97% 95% 93%
Recall 97% 96% 89%
F1-Score 97% 96% 91%

For more details, refer to the Results & Analysis section of the article.

🚀 Getting Started

Prerequisites

  • Python 3.9+
  • Recommended IDE: Visual Studio Code / PyCharm
  • Required packages: Install dependencies using:

Running the Code

  1. Clone the repository:

    git clone https://github.com/rohan-chandrashekar/BoostFex.git
cd repositoryname

  2. Prepare the dataset: Place your dataset in the data/ directory and follow preprocessing steps outlined in scripts/data_preparation.py.

  3. Train the models:

    • For Cyclic Model: 
      python scripts/train_cyclic.py
    • For Bagging Model: 
      python scripts/train_bagging.py

  4. Evaluate performance:

    python scripts/evaluate.py

🤝 Contributing

We welcome contributions to improve the implementation or extend its capabilities. Please submit a pull request or open an issue for discussion.

📝 License

This repository is licensed under the MIT License. See LICENSE for more details.

📬 Contact

For queries, suggestions, or feedback: