Fraud Detection System

A modern, web-based fraud detection system built with Flask and PyTorch Geometric, utilizing a Graph Neural Network (GNN) to predict fraudulent transactions. The system features a compact, user-friendly interface with a vibrant design, allowing users to input transaction details and receive real-time fraud predictions with probability scores.

Table of Contents

• Overview
• Features
• Project Structure
• Installation
• Usage
• Screenshots
• Technologies
• Contributing
• License

Overview

The Fraud Detection System enables users to input transaction details (e.g., zip code, merchant state, transaction amount) and predicts whether a transaction is fraudulent using a pre-trained Graph Attention Network (GAT). The app features a sleek, modern UI with a gradient background, Font Awesome icons, and a compact form for easy interaction. A clear description guides users on how to use the system, and the output includes both a fraud/non-fraud prediction and a probability score.

Features

• Real-Time Fraud Prediction: Uses a GNN model to analyze transaction data and predict fraud.
• Compact & Modern UI: A streamlined form with a gradient background, icons, and hover animations.
• User-Friendly Description: Guides users on input requirements and result interpretation.
• Responsive Design: Optimized for both desktop and mobile devices.
• Robust Error Handling: Validates inputs and provides clear error messages.
• Scalable Architecture: Built with Flask and PyTorch for easy extension.

Project Structure

fraud-detection-system/
├── .gitignore                # Git ignore file for venv, model files, etc.
├── app.py                    # Main Flask application with GNN prediction logic
├── fraud_gnn_model.pth       # Pre-trained GNN model (not tracked in Git)
├── optimal_threshold.txt     # Optimal threshold for fraud prediction (not tracked)
├── templates/
│   └── index.html            # HTML template for the web interface
├── static/
│   └── style.css             # CSS for the modern, compact UI
├── js/
│   └── script.js             # JavaScript for client-side form validation
├── images/
│   └── image.png             # Screenshot or logo for the README

Installation

1. Clone the Repository:

   git clone https://github.com/your-username/fraud-detection-system.git
   cd fraud-detection-system
   

2. Set Up a Virtual Environment:

python -m venv venv
source venv/bin/activate  # On Windows: 

3. Install Dependencies:

pip install flask pandas numpy torch torch-geometric

• Prepare Model Files:
  • Ensure fraud_gnn_model.pth and optimal_threshold.txt are in the project root.
  • If you don’t have these files, train the GNN model or use a dummy optimal_threshold.txt with a value like 0.5 for testing.

4. Run the Application:

python app.py

UsageAccess

• Access the Web Interface:

  • Navigate to http://127.0.0.1:5000 after running the app.

• Enter Transaction Details:Fill out the form with transaction data:

  • Zipcode: e.g., 91750
  • Merchant State: Select from the dropdown (e.g., TX)
  • User Frequency Per Day: e.g., 1
  • Time Difference (Hours): e.g., 16601.95
  • Merchant Category Code: e.g., 5912
  • Merchant City: Select from the dropdown (e.g., Houston)
  • Transaction Amount: e.g., 128.35

Submit the Form:

• Click the "Predict Fraud" button to get a prediction.
• The result will display as “Transaction: Fraud” or “Transaction: Not Fraud” with a fraud probability score.

Handle Errors:

• If an error occurs (e.g., invalid input, missing model files), a clear message will appear in the UI.

Technologies

• Flask: Web framework for the backend.
• PyTorch & PyTorch Geometric: For the GNN-based fraud prediction model.
• Pandas & NumPy: For data processing.
• HTML/CSS/JavaScript: For the frontend interface and validation.
• Font Awesome: For icons in the UI.
• Google Fonts (Poppins): For a modern typography style.