Research On The Construction Of China’s Investor Sentiment Index And Its Influence On Stock Market Forecasting Effect

The constant appearance of financial market anomalies shows that investors in the market are not absolutely rational,and investor sentiment has a great impact on their behavior,which will eventually extend to the whole market.Therefore,the measurement of investor sentiment has become the key,and the partial least square method can extract as much information as possible about investor sentiment,so it is a good method for emotional quantification.With the increasing popularity of the Internet,Baidu is the largest search engine in China and the opinions and emotions of the whole people can be concentrated here to express.Moreover,it has the characteristics of large amount of data,strong timeliness and easy access,so the data extracted from Baidu can more comprehensively and truly reflect investor sentiment.In addition,with the rise of machine learning,because of its ability to self-optimize learning,it also provided a new method for the research of stock market prediction.This paper firstly introduces the relevant theories of investor sentiment and stock market predictions based on extensively reviewing relevant data,and then it constructs the first investor sentiment index by partial least square method based on the four sentiment proxy variables of China’s volatility index,turnover rate,advance-decline line and Shanghai Composite Index return.Based on Baidu’s large database,this paper chooses baidu index data of key words being “bull market” and “bear market” and then borrowed from Mao,Scott and Johan to build the second investor sentiment index.Finally,by using the BP neural network,two prediction models with and without the investor sentiment index are constructed to predict stock market data and the prediction results with and without the sentiment index are compared,so as to achieve the purpose of researching the construction of investor sentiment index and how it affecting the prediction effect of stock market.Through theoretical and empirical research,the following conclusions are found in this paper.(1)There is a significant positive correlation between the two investor sentiment indexes based on the partial least squares method and Baidu index data and the Shanghai Composite Index,and the former has stronger correlation with the Shanghai Composite Index.(2)Because the neural network model has the characteristics of automatically learning from historical data,it has a high prediction accuracy and can effectively predict the closing price of the Shanghai Composite Index and trading volume of Shanghai stock market.Neural network is a very effective method in the field of stock market prediction.(3)After adding the investor sentiment index in the BP neural network,the prediction accuracy of the entire model is significantly improved,which shows that investor sentiment has a positive impact on the prediction of China’s stock market.The addition of the investor sentiment index can make the prediction effect of China’s stock market better,and the investor sentiment index constructed based on the partial least squares method can improve more greatly the prediction accuracy of China’s stock market.(4)After replacing the corresponding data of Shanghai stock market with the closing price of Shenzhen Composite Index and trading volume of Shenzhen stock market,the same research results can still be obtained,which shows that the research conclusions of this paper are universal and reliable.The addition of investor sentiment is conducive to the improvement of stock market forecasting accuracy.