Portfolio Optimization and Backtesting Analysis

This project provides a comprehensive pipeline for portfolio optimization and historical simulation. It is designed to find an optimal portfolio based on historical returns and then simulate its performance over the same period.

Key Features

• Data Caching: Caches historical stock data locally to accelerate subsequent runs. The data is fetched from an external source and saved as generated_output/portfolio_data.pkl.
• Portfolio Optimization: Utilizes Modern Portfolio Theory (MPT) to find the optimal portfolio allocation that maximizes the Sharpe ratio, based on historical data.
• Historical Simulation: Simulates the performance of the optimized portfolio over the historical data period, including periodic rebalancing.
• Advanced Visualization: Generates and saves a suite of plots to the plots/ directory, including:
  • Portfolio Value Over Time
  • Portfolio Drawdown
  • Distribution of Monthly Returns
  • 30-Day Rolling Volatility
  • Asset Correlation Matrix
• Configurability: Key parameters such as the initial investment and risk-free rate can be adjusted in the config.py file.
• Logging: Captures detailed execution information in logs/portfolio_analysis.log.

Project Structure

.
├── plots/                 # Saved visualizations
├── logs/                  # Log files
├── generated_output/      # Directory for cached data
│   └── portfolio_data.pkl # Cached financial data
├── README.md
├── config.py              # Main configuration file
├── main.py                # Main script to run the pipeline
├── portfolio_optimizer.py # Handles portfolio optimization
├── portfolio_simulator.py # Handles historical simulation (backtesting)
└── portfolio_plotter.py   # Handles plotting and visualization

Installation

1. Set up the environment: It is highly recommended to use a virtual environment.

   python -m venv venv
   source venv/bin/activate  # On Windows, use `venv\Scripts\activate`

2. Install dependencies: The project requires several libraries. You can install them using pip:

   pip install pandas numpy matplotlib seaborn scipy

   (Note: A requirements.txt file can be generated with pip freeze > requirements.txt)

How to Run

1. Ensure you have data: Place your cached data file at generated_output/portfolio_data.pkl. This file should be a pandas DataFrame containing the historical price data for your selected assets.
2. (Optional) Customize parameters in config.py. You can adjust settings like the initial investment value.
3. Run the main script from your terminal:

   python main.py

Output

After running the script:

• Console: You will see the final backtest performance metrics printed to the console.
• Plots: All charts will be saved as PNG files in the plots/ directory.
• Logs: The full execution log is saved in the logs/ directory.

Example Visualizations

Below are examples of the plots generated by the analysis.

Portfolio Performance (Equity Curve)

Portfolio Weight Evolution

Drawdown

Rolling Volatility

---

A Note on the Approach

This tool performs optimization based on historical data. The "optimal" portfolio it generates is optimal for that specific historical period only. It is not a prediction of future performance. The simulation is a backtest showing how that single, statically optimized portfolio would have performed.

The results, such as the 100% allocation to a single stock in the example run, are common in historical analyses where one asset has overwhelmingly outperformed others. This does not constitute financial advice, and such a concentrated portfolio would be considered extremely high-risk in a real-world scenario.