An Explanatory Model for the Impact of Investor Sentiment on Stock Markets Integrating XGBoost and SHAP

Abstract

With the widespread popularity of Internet, non-professional individual investors can share information and express their tendency through online forum, exerting a certain influence on the price fluctuations in the stock market. This study has compiled investor sentiment data from the stock forum of Eastmoney.com for the years 2018 to 2022. By integrating this data with the daily return rates of individual stocks in the Shanghai Stock Exchange (SSE), an XGBoost regression model has been established. Utilizing SHAP (Shapley Additive exPlanations) for visualization and analysis, the findings reveal that the daily stock returns exhibit a significant positive correlation with investor positive sentiment and turnover rate, and a significant negative correlation with the number of forum posts. Conversely, there is no significant correlation with metrics such as average number of followers, average reading volume, average number of likes, average net comments, user average influence index, and emotional consistency index within the stock forum. Moreover, in the Chinese SSE market, stocks of different market capitalizations do not display significant differences in sensitivity to investor sentiment. This study contributes novel perspectives and methodologies to the study of investor sentiment, aiding in a more profound comprehension of the behaviors of individual investors participating in stock forum discussions and the consequential impact of such behaviors on the market.