Research on the Nonlinear Impact of Investor Sentiment on Stock Returns Based on Deep Learning and Text Mining

Abstract

This paper focuses on the nonlinear correlation between investor sentiment and stock returns and conducts in-depth research with the aid of deep learning and text mining techniques. First of all, sort out the relevant theoretical cornerstones, covering behavioral finance, market efficiency theory and herd effect theory, to provide theoretical support for the research. Secondly, analyze and study the data context, including the characteristics of text data and stock market data, as well as the necessity of integrated correlation analysis. In terms of extracting investor sentiment indicators, compare the advantages and disadvantages of the sentiment dictionary method, the machine learning method and the deep learning method. Subsequently, an in-depth exploration was conducted on the manifestations, theoretical explanations and implications for investment decisions of the nonlinear impact of investor sentiment on stock returns. Empirical results show that investor sentiment has a significant nonlinear impact on stock returns, and the degree of influence varies among different emotional states. This research provides investors with more accurate sentiment analysis tools to help them predict the trend of the stock market more scientifically and optimize investment decisions.