A Comparative Study of Bayesian Portfolio Optimization: Evidence from AI-Related Stocks

Abstract

This paper conducts a comparative analysis of portfolio optimization methods with a focus on Bayesian approaches, applying them to a dataset of AI-related stocks from the U.S. market. While the classical Markowitz model relies on fixed estimates of return and risk, the Bayesian framework incorporates parameter uncertainty, allowing for more adaptive decision-making. In addition to portfolio construction, the study applies conditional volatility and beta dynamics as a supplementary tool for Bayesian models’ performance analysis, by using the Conditional CAPM model and the DCC-GARCH approach. The performance is evaluated in terms of risk-adjusted returns, particularly the Sharpe ratio, demonstrating the potential advantages of Bayesian optimization in fast-evolving sectors like artificial intelligence. The research finds that although the Markowitz model achieved the highest Sharpe ratio, it also involved the highest concentration risk. Furthermore, the more advanced the Bayesian model, the higher the Sharpe ratio, while conditional volatility and beta levels were simultaneously reduced.