Dynamic Portfolio Optimization Integrating Market Volatility and Ridge Regression

The traditional Markowitz portfolio selection theory provides a framework for balancing returns and risk, but it has a key limitation: it cannot dynamically adjust to changing stock market conditions. In this study, we develop a Markowitz-based portfolio selection model that adjusts stock portfolio weights using a market volatility index reflecting macroeconomic and financial conditions. The model employs Ridge regression, a machine-learning technique, to dynamically estimate the portfolio’s target return, thereby allowing the portfolio to adapt to changing stock market conditions. To evaluate the performance of the proposed model, we applied it to three stock markets—the United States, Germany, and Hong Kong—using historical stock returns and volatility data spanning approximately 15 years, from November 2007 to January 2023. Over this sample period, the proposed model demonstrated superior performance relative to benchmark portfolios across multiple evaluation metrics, including cumulative returns, the Sharpe ratio, and the Sortino ratio. This study empirically demonstrates the effectiveness of the proposed model, which integrates market volatility and Ridge regression, and introduces a multidisciplinary portfolio optimization approach that combines financial engineering, operations research, and machine learning.