FraudShield AI — Intelligent Credit Card Fraud Detection & Transaction Monitoring

FraudShield AI is a web application that helps identify potentially fraudulent credit card transactions using machine learning.

The application uses a trained Weighted XGBoost model to analyze transaction data and generate a fraud probability along with a Fraud or Legitimate result. Each analyzed transaction can be saved and viewed later for monitoring and analysis.

The backend is built with Django and Django REST Framework, with SQLite used for storing transaction records. The frontend is built using HTML, CSS, and JavaScript with a responsive and user-friendly interface.

The project includes a transaction analysis page, monitoring dashboard, transaction history, REST API, sample transaction loader, and detailed developer documentation.

The machine learning model was trained using the ULB Credit Card Fraud Detection dataset, which contains 284,807 transactions, including 492 fraudulent transactions.

The final model achieved:
- ROC-AUC: 0.980
- PR-AUC: 0.883
- Precision: ~0.90
- Recall: ~0.847
- F1-score: ~0.874

The model uses a fixed decision threshold of 0.65 to classify transactions as Fraud or Legitimate.

1. Open the Analyze page.

2. Choose whether to load a sample transaction from the dataset or enter the transaction features manually.

3. If entering a transaction manually, provide the required 30 model features:
Time, V1-V28, and Amount.

4. Click the Analyze Transaction button.

5. FraudShield AI sends the transaction to the backend and passes it through the trained XGBoost model.

6. The system returns a fraud probability and classifies the transaction as Fraud or Legitimate.

7. The analyzed transaction is automatically saved for future reference.

8. Open the Transactions page to search, sort, filter, and view previously analyzed transactions.

9. Open the Dashboard to view transaction statistics, fraud rate, recent activity, and probability charts.

10. Open the Documentation page to learn about the system architecture, machine learning model, and available REST API endpoints.

Python 3.10+
Django
Django REST Framework
XGBoost
scikit-learn
Pandas
NumPy
Joblib
SQLite
HTML5
CSS3
JavaScript
Modern Web Browser