Production-grade, end-to-end quantitative analytics system — VaR, CVaR, Monte Carlo Simulation, ML Signal Generation & Portfolio Optimization across 8 financial stocks (504 trading days).
| Metric | Value |
|---|---|
| 📦 Stocks Analyzed | 8 (504 trading days) |
| 📈 Max Sharpe Ratio | 0.499 |
| 🤖 ML AUC Score | 0.585 (Walk-Forward CV) |
| 🎲 Monte Carlo Paths | 500 (60-day horizon) |
| 💼 Portfolio Return | +37.5% vs SPY +9.5% |
| 🛡️ VaR Confidence | 95% & 99% |
| 🔢 Efficient Frontier | 60 portfolios |
Financial-Risk-Analytics-ML-System/
│
├── main.py ← RUN THIS — Full pipeline orchestrator
├── data_generator.py ← Geometric Brownian Motion market simulation
├── analytics_engine.py ← Risk metrics, portfolio optimization, charts
├── ml_models.py ← ML models, feature engineering, backtest
├── sql_analytics.py ← SQL-style financial queries (CTEs, windows)
├── streamlit_app.py ← Interactive 6-tab Streamlit dashboard
├── requirements.txt ← Python dependencies
├── README.md ← This file
│
├── data/ ← Generated CSV files
└── charts/ ← Generated PNG visualizations
# 1. Clone the repository
git clone https://github.com/dubeypt/Financial-Risk-Analytics-ML-System.git
cd Financial-Risk-Analytics-ML-System
# 2. Install dependencies
pip install -r requirements.txt
# 3. Run full pipeline
python main.py
# 4. Launch interactive dashboard
streamlit run streamlit_app.py| Category | Technology / Method |
|---|---|
| Language | Python 3.10+ |
| Data Manipulation | Pandas, NumPy |
| Statistical Analysis | SciPy (normality tests, z-scores) |
| Machine Learning | Scikit-learn (Random Forest, Isolation Forest) |
| Visualization | Matplotlib, Seaborn, Streamlit |
| Stochastic Modeling | Geometric Brownian Motion (GBM) |
| Risk Analytics | VaR, CVaR, Drawdown, Beta, Alpha |
| Portfolio Theory | Markowitz Mean-Variance, Efficient Frontier |
| Technical Analysis | RSI, MACD, Bollinger Bands, ATR |
| Validation | Walk-Forward Cross-Validation (Time-Series safe) |
| SQL Concepts | CTEs, Window Functions, GROUP BY, CASE WHEN |
- Value at Risk (VaR) at 95% and 99% confidence — daily and portfolio
- Conditional VaR / Expected Shortfall — Basel III standard
- Maximum Drawdown — peak-to-trough analysis
- Volatility Regime Detection — high/low vol regimes
- Sharpe Ratio — return per unit of risk
- Sortino Ratio — downside-adjusted return
- Calmar Ratio — return per unit of drawdown
- Jensen's Alpha — outperformance vs CAPM prediction
- Beta — systematic market risk exposure
- Markowitz Efficient Frontier — 60-point curve
- Maximum Sharpe Portfolio — optimal risk-adjusted allocation
- Minimum Variance Portfolio — lowest risk allocation
- Weight constraints: 2%–40% per stock
- Feature Engineering: 20+ technical indicators as features
- Random Forest Classifier: 5-day return direction prediction
- Walk-Forward Validation: No future data leakage (critical!)
- ROC-AUC evaluation: Industry standard for signal quality
- Isolation Forest: Anomaly detection in price/volume
- Strategy Backtest: ML signal vs Buy-and-Hold
- Monthly returns aggregation (GROUP BY equivalent)
- Rolling Sharpe ranking (Window Functions)
- Volatility regime classification (CASE WHEN)
- Pair correlation CTEs
- P&L attribution (LAG + cumulative SUM)
- Z-score outlier detection (3-sigma events)
- 500 paths, 60-day horizon for GS stock
- Confidence interval bands (5th / 50th / 95th percentile)
- Same GBM model used in Black-Scholes pricing
| File | Description |
|---|---|
charts/01_price_dashboard.png |
5-panel: prices, distributions, correlation, Sharpe, drawdowns |
charts/02_risk_dashboard.png |
VaR/CVaR bars, Monte Carlo, volatility regimes, stress test |
charts/03_portfolio_optimization.png |
Efficient frontier with optimal portfolios |
charts/03_ml_signals.png |
Feature importance, ROC curve, RSI, MACD, anomalies |
charts/04_backtest.png |
ML strategy vs Buy-and-Hold |
data/risk_report.csv |
Full risk metrics for all tickers |
data/sql_monthly_summary.csv |
Monthly aggregated returns |
data/sql_pnl_attribution.csv |
P&L attribution per stock |
data/sql_zscore_outliers.csv |
3-sigma market anomaly events |
reports/final_report.txt |
Complete analytical report |
Aditya Dubey — Data Analyst | Bengaluru, India
