A machine learning-based fraud detection system with a Streamlit web interface for real-time transaction analysis.

This system uses a logistic regression model trained on transaction data to detect potentially fraudulent transactions. The web interface allows users to input transaction details and receive immediate fraud risk assessment.

• Interactive web interface using Streamlit
• Machine learning model for fraud detection
• Support for multiple transaction types
• Real-time prediction capabilities
• Balance tracking for both sender and receiver
• Visual feedback for prediction results

• Python 3.12 or higher
• pip package manager
• Required Python packages (listed in requirements.txt)

1. Clone the repository:

git clone <repository-url>cd Fraud-Detection-System

2. Install required packages:

pip install -r requirements.txt

1. Start the Streamlit web interface:

streamlit run fraud_detection.py

2. Access the application in your web browser at:

http://localhost:8501 

1. Enter transaction details:

   • Select transaction type (PAYMENT, TRANSFER, CASH_OUT, DEPOSIT)
   • Enter transaction amount
   • Input sender's old and new balance
   • Input receiver's old and new balance

2. Click "Predict" to get fraud assessment

3. View results:

   • Green message for legitimate transactions
   • Red warning for potentially fraudulent transactions

Fraud-Detection-System/ ├── fraud_detection.py # Main Streamlit application ├── analysis_model.ipynb # Model training notebook ├── requirements.txt # Project dependencies ├── fraud_detection_model.pkl # Trained model file ├── LICENSE └── README.md 

• Algorithm: Logistic Regression
• Features used:
  • Transaction type
  • Transaction amount
  • Sender's balance (before and after)
  • Receiver's balance (before and after)
• Training data: Financial transaction dataset with labeled fraud cases

If you encounter "command not found" errors:

# Install pipx sudo apt install pipx pipx ensurepath # Install Streamlit globally pipx install streamlit

Feel free to submit issues and enhancement requests.

For any inquiries, please reach out to:

• Email: [email protected]
• GitHub: https://github.com/poncema4