Research On Early Warning Of Stock Market Risk By Investor Sentiment Based On Web Text Mining

Massive Internet financial information has a pivotal position in the financial market,and it has obvious practical value in the excavation of network financial text information.Investors have subjective preferences and cognitive biases in the process of investing,and the market is not completely effective.So asset pricing can not ignore the role of investors'psychological factors.With the emergence of interactive platforms for social networking sites and financial management,investors are more inclined to pay attention to the information of the stock market through convenient ways,and can exchange investment ideas with other investors without being restricted by time and space.If we can effectively combine the Internet financial information to measure investor sentiment and be able to conduct risk warning research,the value brought to the stock market practice is particularly significant.Under the framework of modern asset theory,this paper firstly analyses the influence mechanism of investor sentiment on stock market risk from the perspective of psychology and external environment.Then,this article takes 554,186 posts of stock bar about Shanghai securities composite index in the eastmoney website from March 17,2017 to May 3,2017 as the research object,to comprehensively portray the emotional characteristics of investors and apply the dictionary-based sentiment analysis method to realize financial text mining and emotion recognition.At the same time,it designs a stock market risk early warning system and applies market performance indicators and investor sentiment indicators to the SVM model with the radial function as the radial basis function,the penalty coefficient of 4,and the parameter of 2^-5,and empirically examines the investor sentiment in the stock market.With the accuracy of early warning,precision,recall and F1 index to scientifically assess the early warning capability,the accuracy of back-testing of investor sentiment reached 90.48%,which improved the early warning sensitivity of unbalanced data and basically met the dynamic monitoring demand of stock market risk.In order to resolutely fight the tough battle to prevent and resolve major risks,this article deeply integrates emerging text mining technology and creatively realizes the comprehensive evaluation of investor sentiment from the perspective of 3D sentiment,investor attention and investor interaction.On the other hand,it combines with market performance and SVM in machine learning because we try to find the best risk warning model.Finally,conclusions are as follows:Firstly,there is a trading-day effect in investor posting.Secondly,investor sentiment based on web text mining shows short-term aging characteristics.Thirdly,the relationship between investor sentiment and stock market performance contained in the online financial text is linked,and the risk level of the stock market can be predicted according to the investor sentiment under the network information.Fourthly,the SVM early warning model that introduces investor sentiment has superior early warning performance for stock market risk.The model not only improves the results of risk classification under unbalanced data sets,but also optimizes the early warning effect based on the model of market performance indicator.