🛡️ FraudGuard AI: Real-Time Credit Card Fraud Detection

FraudGuard AI is an end-to-end Machine Learning application designed to detect fraudulent credit card transactions in real-time. It addresses the challenge of extreme class imbalance (only 0.17% fraud cases) using a robust XGBoost classifier and provides a user-friendly interface via Streamlit.

---

🚀 Live Demo

Click here to use the Live App

---

🧠 Key Features

• High-Performance Model: Built using XGBoost (Extreme Gradient Boosting), optimized for tabular data.
• Imbalance Handling: Utilizes scale_pos_weight to mathematically penalize missing fraud cases, achieving high Recall.
• Real-Time Inference: Instant predictions with a probability risk score.
• Production Ready: Modular code structure, pipeline-based preprocessing, and containerized with Docker.
• Interactive UI: A dark-mode dashboard to simulate transaction scenarios and visualize risk factors.

---

🛠️ Tech Stack

• Language: Python 3.11
• Machine Learning: Scikit-Learn, XGBoost, Joblib
• Data Processing: Pandas, NumPy
• Deployment: Streamlit Cloud, Docker

---

💻 How to Run Locally

You can run this project on your local machine using Python or Docker.

Option 1: Using Python (Recommended)

1. Clone the repository:

   git clone https://github.com/DhashubhanKumar/fraud-detection-live.git
   cd fraud-detection-live

2. Create a virtual environment: It is recommended to use Python 3.11 to avoid compatibility issues.

   # Windows
   py -3.11 -m venv venv
   .\venv\Scripts\Activate
   
   # Mac/Linux
   python3.11 -m venv venv
   source venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

4. Run the App:

   streamlit run app.py

   The app will open in your browser at http://localhost:8501.

Option 2: Using Docker 🐳

If you have Docker installed, you can run the app in a container without installing Python libraries.

1. Build the Image:

   docker build -t fraud-app .

2. Run the Container:

   docker run -p 8501:8501 fraud-app

   Access the app at http://localhost:8501.

---

🧪 How to Simulate Fraud

The model is trained on real patterns. To see the model flag a transaction as FRAUD in the UI, set the sliders to the following values (which mimic a typical mathematical fraud signature):

1. V14: Set to Negative (e.g., -5.0)
2. V4: Set to Positive (e.g., 5.0)
3. V11: Set to Positive (e.g., 4.0)

Click "Analyze Transaction" and watch the alert turn RED. 🚨

---

📊 Model Metrics

The model was evaluated on the unseen test set of the Kaggle Credit Card Fraud Dataset.

Metric	Score	Description
ROC-AUC	0.97+	Excellent ability to distinguish between Fraud and Safe.
Recall	High	Prioritizes catching thieves over avoiding false alarms.

---

👨‍💻 Author

Dhashubhan Kumar
Aspiring GenAI Engineer

---

Note: The dataset features (V1-V28) are PCA-transformed for privacy. This project simulates the backend processing of a banking security module.