A Study on Machine Learning-Enhanced Efficient Frontiers in Multifactor Models

Abstract

Traditional multi-factor models and Markowitz’s efficient frontier theory struggle to adapt to complex financial markets, making machine learning a crucial tool for expanding the efficient frontier of investment portfolios. This paper systematically reviews the research landscape in this field, delineates its developmental stages, and summarizes the current applications and pros and cons of various machine learning algorithms across four dimensions: factor discovery, risk measurement, portfolio optimization, and constraint refinement. The study finds that the current field suffers from issues such as a lack of underlying theory, a fragmented research framework, the presence of pseudo-factors, and overfitting. Based on these findings, future research directions are projected, and it is proposed that the theoretical framework should be strengthened, causal and robust techniques should be integrated, and localized research should be conducted, with the aim of providing guidance for both theoretical exploration and practical applications in this field.