Advances in International Finance

This study analyzes the fluctuations in investor sentiment in the Chinese stock market and constructs a comprehensive investor sentiment index that includes macroeconomic indicators and sentiment analysis. Machine learning models such as LSTM and BiLSTM are utilized to predict stock returns. The results indicate that the model can quickly reflect sentiment changes during market volatility, providing accurate forecasts for investors and enhancing the understanding of the relationship between market sentiment and stock returns, as well as the scientific nature of investment decisions.

Investor sentiment index; Machine learning; Stock returns