FinRisk Analytics

Volatility Forecasting & Value-at-Risk Modeling Using ML & Econometrics

---

🎯 Overview

FinRisk Analytics is an institutional-grade risk modeling platform that combines classical econometric methods (GARCH, EGARCH) with modern machine learning (LSTM, XGBoost) to forecast volatility and calculate Value-at-Risk across multiple asset classes.

Key Features:

• 📊 Multi-model volatility forecasting (GARCH, EGARCH, LSTM)
• ⚠️ Value-at-Risk calculation (5 methods)
• 📈 Expected Shortfall (CVaR) analysis
• 🔍 Statistical backtesting (Kupiec test)
• 🔗 Correlation regime detection
• 📱 Interactive Streamlit dashboard
• 📉 Crisis period stress testing

---

🏗️ Project Structure

finrisk-analytics/
├── configs/              # Configuration files
├── data/                 # Data storage
│   ├── raw/             # Raw market data
│   ├── processed/       # Cleaned data
│   └── features/        # Engineered features
├── notebooks/           # Jupyter notebooks (5 total)
├── src/                 # Source code
│   ├── data/           # Data handling
│   ├── models/         # GARCH & LSTM models
│   ├── risk/           # VaR calculation & backtesting
│   ├── correlation/    # Correlation analysis
│   └── utils/          # Utilities
├── dashboard/          # Streamlit dashboard
├── results/            # Outputs
│   ├── plots/         # Visualizations
│   ├── reports/       # Risk reports
│   └── models/        # Trained models
└── tests/             # Unit tests

---

🚀 Quick Start

1. Clone Repository

git clone https://github.com/aniktahabilder/finrisk-analytics.git
cd finrisk-analytics

2. Set Up Environment

# Create virtual environment
python3 -m venv venv

# Activate (Linux/Mac)
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

3. Configure Settings

Edit configs/config.yaml to customize:

• Date ranges
• Asset universe
• Model parameters
• VaR confidence levels

4. Run Analysis

Option A: Jupyter Notebooks

jupyter notebook

# Run notebooks in order:
# 01_volatility_eda.ipynb
# 02_garch_models.ipynb
# 03_lstm_volatility.ipynb
# 04_var_calculation.ipynb
# 05_portfolio_risk.ipynb

Option B: Interactive Dashboard

streamlit run dashboard/app.py

---

📊 Asset Universe

15 Assets across 4 Classes:

Category	Assets	Purpose
Equities	SPY, XLK, XLF, XLV, XLE, XLI	Market + sector risk
Fixed Income	TLT, IEF	Duration risk
Alternatives	GLD, BTC, ETH, SOL	Diversification risk
Macro	VIX	Volatility indicator

Deep-dive case studies: SPY, BTC, ETH, TLT

---

🔧 Methodology

Volatility Models

Classical Econometrics:

• GARCH(1,1) - Baseline volatility clustering
• EGARCH - Asymmetric volatility (leverage effect)
• GJR-GARCH - Captures negative shock impact

Machine Learning:

• LSTM - Sequential time-series patterns
• LSTM-GARCH Hybrid - Combined approach
• XGBoost - Feature-based volatility

Value-at-Risk Methods

1. Historical Simulation - Empirical distribution
2. Parametric (Normal) - Gaussian assumption
3. Parametric (Student-t) - Fat tails
4. Cornish-Fisher - Skewness/kurtosis adjusted
5. Filtered Historical Simulation - GARCH + historical

Risk Metrics

• Value-at-Risk (95%, 99% confidence)
• Expected Shortfall (CVaR)
• Rolling correlations
• Regime-dependent correlations
• Portfolio VaR (multi-asset)

Backtesting

• VaR violation counting
• Kupiec test (statistical validation)
• Traffic light approach (Basel framework)
• Crisis period analysis (March 2020, 2022)

---

📈 Target Metrics

Metric	Target
Volatility Forecast RMSE	< 2%
VaR Coverage (95%)	4-6% violations
Kupiec Test	Pass at 5%
Expected Shortfall Error	< 10%
Dashboard Load Time	< 3 seconds

---

📱 Dashboard Features

Interactive Streamlit App:

• Real-time volatility charts
• VaR breach timeline visualization
• Correlation heatmaps (by regime)
• Portfolio risk calculator
• Model comparison view
• Crisis scenario analysis

Launch:

streamlit run dashboard/app.py

---

🔗 Related Projects

This is part of a three-project portfolio:

1. Market Intelligence ML - Price prediction
2. FinRisk Analytics ← You are here
3. AlphaRL Portfolio - RL optimization

Integration: This project provides volatility forecasts and VaR estimates that feed into Project 3's RL agent for risk-aware portfolio optimization.

---

📚 Dependencies

Core:

• pandas, numpy, scipy
• scikit-learn, xgboost, tensorflow

Econometrics:

• arch (GARCH models)
• statsmodels

Visualization:

• matplotlib, seaborn, plotly
• streamlit (dashboard)

Financial Data:

• yfinance, ccxt

See requirements.txt for complete list.

---

🎯 Use Cases

Risk Management:

• Portfolio risk monitoring
• VaR calculation for regulatory compliance
• Stress testing under crisis scenarios
• Early warning system for volatility spikes

Quantitative Research:

• Volatility model comparison
• Risk factor attribution
• Correlation regime analysis

Portfolio Management:

• Risk budgeting
• Dynamic hedging strategies
• Capital allocation decisions

---

🧪 Testing

# Run unit tests
pytest tests/

# Run with coverage
pytest --cov=src tests/

---

📝 To-Do

• Implement data fetching pipeline
• Fit GARCH models on all assets
• Train LSTM volatility models
• Calculate VaR using all methods
• Run Kupiec backtesting
• Analyze correlation regimes
• Build interactive dashboard
• Generate risk reports
• Crisis period analysis
• Write technical documentation

---

🤝 Contributing

This is a personal portfolio project. Feedback and suggestions are welcome!

---

📄 License

MIT License - see LICENSE file.

---

👤 Author

Anik Tahabilder

• PhD Student, Computer Science, Wayne State University
• Research: Multimodal AI, Blockchain Security, Quantitative Finance
• GitHub: @aniktahabilder

---

🙏 Acknowledgments

• GARCH methodology from Engle (2001) and Bollerslev (1986)
• VaR backtesting framework based on Basel Committee guidelines
• Built as part of comprehensive ML/Finance portfolio

---

📞 Contact

For questions or collaboration:

• GitHub: aniktahabilder
• Email: [your-email]

---

⭐ If you find this project useful, please consider giving it a star!