Скачать или смотреть Advanced Portfolio Optimizer in Python: Cost-Aware Markowitz & Optuna Hyperparameter Tuning

Ever wondered how much transaction costs and portfolio drift really impact your returns? In this video, I unveil an advanced portfolio optimization tool built with Python and Streamlit that goes beyond textbook theory to provide realistic, robust backtests.
This application performs a rigorous, point-in-time, walk-forward analysis, comparing a modern Cost-Aware Markowitz strategy against the Classic Markowitz model and an Equal-Weight benchmark. See for yourself how accounting for real-world frictions can dramatically change performance outcomes.
The highlight of this project is the new Automated Parameter Optimization feature powered by Optuna. Watch as the application intelligently searches for the optimal risk management parameters (stop-loss, take-profit, etc.) on a training dataset and then validates them on unseen data to prevent overfitting.
Key Features of the Optimizer:
🔹 Survivorship Bias-Free Backtesting: The simulation dynamically selects a tradable universe of stocks at each step, naturally handling IPOs and delistings to avoid survivorship bias.
🔹 Realistic, Drift-Aware Transaction Costs: Costs are calculated based on the portfolio's actual "drifted" weights, not the previous target weights, providing a much more accurate picture of turnover expenses.
🔹 Unbiased, Point-in-Time Logic: All calculations are performed strictly using information available at that moment in time, ensuring there is no lookahead bias.
🔹 Automated Hyperparameter Tuning with Optuna: A powerful new feature that automatically finds the best risk parameters and rebalance window, saving you from endless manual guesswork.
🔹 Fair & Flexible Risk Management: Apply portfolio-level or individual stock-level stop-loss/take-profit rules equally across all strategies for a true apples-to-apples comparison.