Research On The Influence Of Microblogs Sentiment On Stock Market Using A Multi-strategy-based Classification Approach

With the rapid development of Internet technology,social networking is more and more widely applicable.Not only has Social networking become the main channel for people to communicate with each other,but also it has become the main channel for people to participate in social public discussion and vent their thoughts and emotions.In-depth analysis and excavation of information released by the public on the social platform has become an effective way or channel for timely understanding and grasping public sentiment.In recent years for different application areas,how to make full use of information on the social platform,timely analysis and grasp the relevant development trends,not only more and more enterprises and government departments concerned,but also has become one of the hot research issues.Weibo in China is a social platform with wide audience and high openness.Information on Weibo can be better to reflect from a side the status of economic and social development.In China there have been some scholars in the relevant areas to carry out a number of microblogging information analysis and application of research and achieved some preliminary results.Based on the study of analysis method microblogs sentiment,the relationship between microblogs sentiment and the fluctuation of stock market in China is discussed.In the theoretical research,first of all,the concept of public sentiment,microblogs sentiment and sentiment index are defined.Secondly,based on the existing research results,the construction process of microblogs sentiment index based on multi-strategy classification method is put forward.From the classification process,the method called two-step three categories is used,namely,the subjective and objective classification of all texts and the appraisable classification of subjective texts.Finally,all texts are classified into three categories which are neutral text or objective text,active text and negative text.From the classification method,the vector space model and the emotion characteristic model respectively are proposed to construct.In this two models,the three kinds of classification algorithms,SVM,Logistic Regression and naive Bayesian,were used to classify the text and the optimal voting of 6 kinds of classification results was designed.After that the methods applied in the classification process and the methods of constructing microblogs sentiment index are elaborated.In practical application research,firstly,the data acquired on Sina Weibo is token as the research object,the process of data acquisition and preprocessing and is expatiated and the construction of Sina Weibo stock investment sentiment index is completed with proposed methods.Secondly,the relevant dependent variables and independent variables in the stock market forecasting study are chosen and the calculated weibo stock investment sentiment is considered as the independent variable.On the one hand,the relationship between weibo stock investment sentiment index and volatility of selected stock market is studied.Granger causality test is used to select the weibo stock investment sentiment index which is related to stock market fluctuation.On the other hand,the influence of weibo stock investment sentiment index on the accuracy of stock market forecasting is studied.The result shows that the Bullishness index and the simple emotional index after first-order difference are correlated with the fluctuation of the stock market and the forecasting model of weibo stock investment sentiment index and the stock market volume index has the highest forecast accuracy.