Fraud Detection Explorer

An interactive Streamlit dashboard for exploring, analyzing, and detecting fraudulent transactions using XGBoost and SMOTE.

Overview

Fraudulent transactions are rare but costly. This app provides an interactive environment to:

• Perform Exploratory Data Analysis (EDA) on transaction data.
• Visualize class imbalance, feature distributions, and correlations.
• Train and evaluate an XGBoost model with SMOTE to handle imbalanced data.
• Predict fraud probabilities for new transactions.
• Interpret predictions using SHAP (SHapley Additive exPlanations).

All wrapped into an intuitive Streamlit dashboard that helps data scientists, analysts, and fraud investigators explore their datasets in depth.

---

Features

• EDA Dashboard:
  Explore missing values, outliers, feature distributions, and correlations.

• Model Training:
  Train an XGBoost model directly in the app with adjustable hyperparameters.

• Performance Metrics:
  View confusion matrix, ROC curve, and precision-recall curves interactively.

• Prediction Mode:
  Upload new transactions and predict fraud probabilities instantly.

• Explainability:
  Visualize SHAP summary plots to understand which features drive fraud predictions.

---

🧩 Project Structure

fraud-detection/ │ ├── app.py # Streamlit dashboard ├── model.py # Model training and SHAP logic ├── utils.py # Helper utilities (EDA plots, etc.) ├── data/ │ └── creditcard.csv # Sample dataset (if available) ├── trained_model.joblib # Saved model after training (optional) ├── requirements.txt # Python dependencies └── README.md # You are here 🚀

⚙️ Installation & Setup

1. Clone this repository:

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

2.Create a virtual environment (recommended):

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

3.Install dependencies:

pip install -r requirements.txt

4.Run the app:

streamlit run app.py

📊 Example Screenshots

🧠 Tech Stack

Frontend: Streamlit

Modeling: XGBoost, SMOTE (imbalanced-learn)

EDA & Visualization: Pandas, Matplotlib, Seaborn

Explainability: SHAP

💡 Future Improvements

Add time-based or geolocation fraud pattern analysis

Integrate live data ingestion API

Deploy to Streamlit Cloud or Hugging Face Spaces

Add a model performance comparison tab (XGBoost vs. Logistic Regression)