Investor Sentiment Index Construction And Empirical Research Based On Sentiment Analysis

Stock market forecasting has always been a research hotspot,however,due to the complexity of the stock market,trends are hard to predict.The development of behavioral finance shows that investors in the market are not completely “reasonable people”,the psychological factors such as bias,mood,emotion,and preference in the cognitive process make it impossible for investors to make decisions in a rational way,in fact,fallacies often appear in their behavior.With the rapid development and popularization of the Internet,more and more investors search information,exchange opinions and express emotions through the Internet.The Internet provides a huge amount of data for analyzing investor sentiment.The development of technologies such as text mining and machine learning has enabled us to deal with large-scale Internet data,and providing a new method for measuring investor sentiment.With theories,data,and methods available,it is possible to use sentiment analysis techniques to study the relationship between investor sentiment and the stock market.This thesis is based on behavioral finance theory,using sentiment analysis technology to analyze investor sentiment,and study the relationship between investor sentiment and stock price.This thesis have captured a large number of comments on the Shanghai Composite Index and used machine learning techniques to extract investor sentiment to construct sentiment indicators.After the completion of the sentiment indicator construction,this thesis analyzed the correlation between these indicators and the price,volatility,volume,etc.of the Shanghai Composite Index.Finally,established a forecast model for the Shanghai Composite Index by combining historical price data with sentiment indicators.This thesis attempts to use transfer learning to improve the accuracy of the sentiment classification model.In addition,this thesis proposes a integrated research framework.Under this framework,we can use investor sentiment indicators to develop a stock market forecasting system with practical value.This is also the innovation of this thesis in engineering practice.In this thesis,the correlation coefficient between the Bullishness Moving Average Index and the Shanghai Composite Index is 0.631518,and the accuracy of the prediction model is 71.3%.This shows that the investor sentiment on the Internet is closely related to the trend of the stock market,and it is feasible to use investor sentiment to predict the trend of the stock market.