Machine Learning Approaches to Picking A-Shares Stocks: A Comparative Analysis

This study explores the integration of advanced machine learning (ML) techniques and large language models (LLMs) in financial modeling, focusing on the Chinese stock market. It introduces the ChatGPT Score, an LLM-driven sentiment analysis factor, and compares the traditional Fama-French five-factor (FF5) model with its augmented version, FF5+ChatGPT Score. The research evaluates linear regression models against ML models, such as Random Forests, Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), and Category Boosting (CatBoost), within five- and six-factor frameworks. Empirical results show that the ChatGPT Score outperforms traditional sentiment tools like SnowNLP and improves the predictive accuracy of the FF5 model. Additionally, CatBoost and Random Forests demonstrate strong portfolio management capabilities. Statistical validation through retrospective analysis confirms the effectiveness of the models, while industry feedback highlights their practical value in investment strategies. However, the study acknowledges the limitations of current models and recommends future research on deep learning techniques to improve financial market analysis and predictive accuracy.