생성형 인공지능을 활용한 투자 포트폴리오 형성에 관한 실증 분석 연구

Purpose - The purpose of this study is to empirically verify whether generative artificial intelligence (AI) can be used as a tool to positively influence investors’ investment strategies. Design/methodology/approach - For this purpose, we received recommendations from GPT-4o for stocks of different sizes and weights among the large and liquid securities that make up the CSI 300 Index, formed five portfolios through mean-variance optimization, and compared and analyzed the changes in cumulative returns. Findings - The empirical results showed that the cumulative returns of portfolios based on stocks rated as excellent by GPT-4o were better than those of the market index, CSI300, regardless of the portfolio composition. In addition, the cumulative return performance estimated by the portfolios weighted by GPT-4o did not differ from the cumulative returns of the portfolios formed by applying the financial mean-variance optimization theory. Research implications or originality - The results of this empirical analysis show that generative AI models have ample potential to be used as a tool to have a positive effect on investment strategies, such as portfolio construction, beyond simple stock selection.