Taking the main component stocks of the CSI 300 index as the research object, the specific impact of investor sentiment on the stock market was explored. Firstly, four different methods for constructing investor sentiment factors were proposed, and the most effective method by analyzing the sentiment information from the East Money stock bar was determined. Next, the constructed investor sentiment index was combined with traditional financial indicators, and the Fama-French four-factor model and random forest model were used to conduct in-depth analysis of the impact of sentiment factors on stock returns. The results show that investor sentiment factors have a significant predictive ability for changes in stock returns.