📊 Quantitative Portfolio Analytics Platform

Production-grade portfolio optimization and risk analysis system. Processes 24,500+ days of market data across 25+ equities and cryptocurrencies using modern portfolio theory.

---

Features

Portfolio Optimization Engine

• Mean-variance optimization (maximize Sharpe ratio / minimize volatility)
• Ledoit-Wolf covariance shrinkage with matrix regularization
• Multi-solver support (ECOS/SCS/OSQP) with automatic fallback
• Numerical stability handling: eigenvalue analysis, positive semi-definite enforcement

Data Infrastructure

• DuckDB analytics database for high-performance time-series analysis
• Yahoo Finance API integration with automated validation
• Robust data pipeline with error handling and outlier detection
• Efficient column-oriented storage with SQL interface

Risk Analytics

• Sharpe ratio, volatility, maximum drawdown analysis
• Rolling performance windows (60-day metrics)
• Portfolio risk attribution
• Strategy comparison and benchmark analysis

Interactive Dashboard

• Real-time portfolio visualization with Plotly
• Multiple optimization strategies and asset classes
• Dynamic weight allocation display
• Performance tracking with cumulative returns

---

Tech Stack

Python • pandas • NumPy • PyPortfolioOpt • DuckDB • Streamlit • Plotly • yfinance

---

Quick Start

# Setup
git clone https://github.com/leaalonzo/quant-portfolio-analytics-dev.git
cd quant-portfolio-analytics-dev
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

# Run pipeline
python -m utils.fetch_all_data           # Fetch market data
python scripts/build_factors.py          # Process data
python scripts/run_backtest.py           # Run backtests
streamlit run app/app.py                 # Launch dashboard

---

Project Structure

├── app/app.py                    # Streamlit dashboard
├── data/                         # CSV & DuckDB files
├── scripts/
│   ├── build_factors.py         # Data processing
│   └── run_backtest.py          # Portfolio backtesting
├── utils/
│   ├── optimize.py              # Portfolio optimization
│   ├── backtest.py              # Portfolio construction
│   └── risk.py                  # Risk analytics
│   └── fetch_all_data.py        # Fetch raw data from API
└── requirements.txt

---

Optimization Implementation

Mean-Variance Framework

• Max Sharpe: Maximize risk-adjusted returns using PyPortfolioOpt
• Min Volatility: Minimize portfolio variance
• Covariance: Ledoit-Wolf shrinkage for robust estimation
• Constraints: Position limits (0-50% per asset)

Numerical Robustness

# Covariance regularization
S = returns.cov() * 252
S = S + np.eye(len(S)) * 0.1  # 10% diagonal loading

# Matrix symmetry enforcement
S = (S + S.T) / 2

# Multi-solver fallback
solvers = ['ECOS', 'SCS', 'OSQP']
for solver in solvers:
    try:
        ef = EfficientFrontier(mu, S, solver=solver)
        # Optimization succeeds
    except:
        # Try next solver

Data Cleaning Pipeline

• Infinity value replacement
• NaN handling (forward-fill → backward-fill → zero-fill)
• Extreme value clipping (-50% to +100% daily returns)
• Zero variance detection and removal

---

Key Features

Robust Optimization

• Automatic solver selection and fallback mechanism
• Handles numerical instability in crypto data
• Equal-weight fallback when optimization fails
• Performance validation (detects invalid results)

Data Quality Management

• Dynamic NaN threshold (80% for equities, 50% for crypto)
• Asset-class specific data handling
• Comprehensive validation checks
• Diagnostic logging and error reporting

Portfolio Strategies

• Long-only and long-short portfolios
• Equal-weight and optimized allocations
• Multiple rebalancing approaches
• Strategy-specific asset selection

---

Sample Results

Equity Portfolio - Max Sharpe
├── Expected Return: 12.4%
├── Volatility: 15.8%
├── Sharpe Ratio: 0.66
└── Assets: 15 equities

Crypto Portfolio - Min Volatility
├── Expected Return: 28.7%
├── Volatility: 42.1%
├── Sharpe Ratio: 0.63
└── Assets: 7 cryptocurrencies

---

System Architecture

Data Collection → DuckDB Storage → Portfolio Backtesting → Optimization → Dashboard
     ↓                ↓                    ↓                    ↓            ↓
  yfinance         SQL DB          Equal-weight         PyPortfolioOpt   Streamlit
                                   Long/Short           Ledoit-Wolf      Plotly

---

Performance Metrics

Returns

• Cumulative returns
• Annualized returns
• Rolling performance

Risk

• Volatility (annualized)
• Maximum drawdown
• Downside deviation

Risk-Adjusted

• Sharpe ratio
• Sortino ratio
• Rolling Sharpe

---

Deployment

Local

streamlit run app/app.py

Docker

docker build -t quant-portfolio .
docker run -p 8501:8501 quant-portfolio

Cloud: Streamlit Cloud, AWS, GCP

---

Database Queries

-- Portfolio performance
SELECT date, cumulative
FROM portfolio_results
WHERE asset_type = 'Equity';

-- Holdings analysis
SELECT tickers, COUNT(*) as appearances
FROM portfolio_holdings
GROUP BY tickers
ORDER BY appearances DESC;

---

Author

---

Disclaimer

This is a personal educational project developed independently.

---

License

MIT License

---

⭐ Star this repo if you find it useful!