Credit Card Fraud Detection

🎯 Objective
Detect fraudulent credit card transactions using machine learning, featuring an interactive Streamlit web app for demo and exploration.

---

🚀 Features

• Tabular data from Kaggle (284K transactions; 492 fraud)
• Models implemented:
  • Random Forest (primary saved model)
  • Logistic Regression & XGBoost (for experimentation)
• Imbalanced dataset handling via scaling & model tuning
• Interactive Streamlit app with:
  • Select a simulated user transaction
  • Show fraud probability and decision
  • Option to display a random fraud case
  • Configurable fraud threshold (moving towards)

📸 App Screenshots

🏠 Homepage

🔍 Prediction Output

---

📁 Repo Structure

credit-card-fraud-detection/ ├── README.md ├── LICENSE ├── .gitignore ├── requirements.txt ├── data/ │ ├── creditcard.csv │ └── fraud_data_simulated.csv ├── models/ │ ├── fraud_model.pkl │ └── scaler.pkl ├── notebooks/ │ ├── 01_EDA.ipynb │ └── 02_Model_Training.ipynb ├── scripts/ │ └── add_fake_fields.py ├── src/ │ └── model_training.py ├── app.py

---

💻 Setup Instructions

1. Clone and install dependencies.

   git clone <YOUR_REPO_URL>
   cd credit-card-fraud-detection
   pip install -r requirements.txt
   

2. Download creditcard.csv from [Kaggle dataset].

Place it in the data/ folder.

Generate simulated data

python scripts/add_fake_fields.py
Train & save model

python src/model_training.py
Run web app

streamlit run app.py

3. 📈 Usage

• Use the dropdown to select a "user"
• Or check “Show me a fraud case” to randomly sample a fraud

The app shows:

--> Customer details --> Fraud probability --> Decision (Legit vs Fraud)

4. ⚙️ Future Improvements

• Add SMOTE/ADASYN for class imbalance
• Use GridSearchCV or Optuna for hyperparameter tuning
• Add cross-validation & evaluation metrics
• Track experiments with MLflow / Weights & Biases
• Deploy publicly via Streamlit Cloud or Hugging Face Spaces
• Improve UI: add charts, threshold slider, download option

5. 📄 License This project is open-source under the MIT License. See LICENSE for details.

6. 🛠️ Built With --> Python, pandas, scikit-learn, XGBoost --> Streamlit for app UI --> joblib for model persistence