The Impact of Investor Sentiment on Stock Returns Based on Machine Learning and Deep Learning Methods

The popularity of artificial intelligence, as demonstrated by ChatGPT, continues to drive the remarkable growth of AI-related concept stocks in the market and has a significant influence on the overall performance of technology stocks. However, multiple factors affect the stock prices of technology companies. This study seeks to examine how investor sentiment impacts the stock returns of technology companies using the Fama-French three-factor model. We employ principal component analysis to conduct factor analysis, create an investor sentiment factor, and propose the utilization of a TiDE time series model based on a multilayer perceptron (MLP) to forecast stock returns. We progressively introduce four indicators, namely book-to-market ratio, market return, total market value, and the investor sentiment factor, and observe a substantial improvement in the accuracy of stock return predictions. Additionally, when investors feel more positive about a particular stock or the market as a whole, there tends to be an increase in stock returns. Conversely, when investor sentiment is negative, there tends to be a decrease in stock returns. Comparing the TiDE model with machine learning methods like Random Forest and Gradient Boosting, as well as deep learning methods such as LSTM and Transformer, we find that the TiDE model enhances prediction accuracy and reduces the disparity between predicted and actual values in time series forecasting tasks. On the one hand, this study helps investors better understand the impact of investor sentiment on the stock prices of technology companies in China’s financial markets. On the other hand, it provides empirical research evidence for applying artificial intelligence in the financial field. In the future, this research result can be useful in applying to a wider range of stock markets and other financial fields.