Research On The Topics Of Stock Comments Based On Sentiment Analysis

In the era of rapid development of the Internet,there are more and more platforms for netizens to express their opinions,and gradually become the main carrier of online public opinion.At the same time,China’s economy shows a trend of gradual and steady growth,the national disposable funds continue to increase,more and more shareholders through the Internet to participate in investment and express their views.To some extent,the comments of some shareholders affect the management decisions of relevant management departments and the purchase decisions of other investors,which has an important impact on China’s stock market.However,China’s stock market is still immature,and the relevant management departments have not yet been able to carry out more effective and comprehensive governance.Therefore,this dissertation starts with the information data of stock comments,explores the emotions of shareholders and guides public opinion events to develop in a positive direction,which is of great significance for the relevant management departments to carry out efficient and comprehensive governance.Based on the statistics of website activity,this dissertation studies the stock comment data of Kweichow Moutai’s stock bar on Oriental Fortune website platform.Based on the general sentiment dictionary of How Net and the word characteristics in the financial field,the exclusive sentiment dictionary of Kweichow Moutai’s stock comment field is constructed using Word2 Vec.In the process of analyzing the emotional tendency of text data,different weight values are assigned to adverbs of different levels of degree by calculating accuracy,and the emotional value of each comment is finally calculated,and the number of different types of emotional text is displayed by pie chart.Then,LDA(potential Dirichlet distribution)topic generation model is used to extract the positive and negative texts under sentiment classification,and the optimal number of topics is determined by calculating the confusion degree.The feature words of different topics are displayed,and different topics are summarized according to the implied semantics.Then,with shareholders as nodes and comments between shareholders as edges,the Gephi visualization software is used to construct the topic clustering map,and high influential shareholders are displayed according to the centrality.Finally,from the aspects of emotional guidance,information content,high-influence shareholders and so on,decision-making suggestions are put forward for relevant management departments to carry out effective public opinion control.