This project implements a dynamic, regime-aware portfolio risk modeling framework as per the thesis requirements. It integrates statistical validation, time-series forecasting, and machine learning to measure downside risk.
- FX-Adjusted Data: Live NSE data (Reliance, TCS, INFY) adjusted for USD/INR.
- Coherent Risk Metrics: Historical and Parametric Value at Risk (VaR) & Expected Shortfall (ES).
- Regime Detection: K-Means Clustering to identify High/Low volatility market states.
- Volatility Forecasting: ARIMA (1,1,1) model for time-series risk prediction.
- ML Classification: Random Forest Classifier for risk state prediction.
- Stress Testing: Scenario analysis for market crashes and currency shocks.
- Visualization: High-resolution dashboard for thesis inclusion.
- Ensure Python 3.10+ is installed.
- Install dependencies:
pip install yfinance pandas numpy matplotlib seaborn scikit-learn statsmodels - Run the master pipeline:
python main.py
/src: Modular Python scripts for each methodology./data: CSV files containing raw and processed analytical data./outputs: Professional charts and ML classification reports.