Research On The Evolution Of Net-Mediated Public Sentiment Of Online Financial Violations Based On Text Analysis

With the development of communication and network technology in the era of big data,the Internet industry develops rapidly,which is reflected in various aspects such as social networking platform,Internet financial platform,and has an impact on the traditional social life style.In this context,the information dissemination mechanism is more open and convenient,netizens can express and disseminate the opinions and attitudes of specific events at any time.Therefore,the research on net-mediated public sentiment is increasingly concerned by government and relevant scholars.Internet finance industry has developed rapidly in China,but some enterprises do not pay enough attention to the control of risk,which makes Internet finance have great risk.The net-mediated public sentiment of financial violations spreads on the Internet may cause a lot of social attention,even affect the social and cultural environment.The huge amount of data in the form of text can be used as information source in the supervision of financial risk.By using the text analysis method,public opinion can be accurately identified,which is conducive to improving the emergency response mechanism of financial risk.The dissemination of net-mediated public sentiment is a dynamic development cycle.Based on life cycle theory,the existing researches have divided the development stages of public opinion.The influencing factors of the public opinion development include enterprise,government,netizens and media four aspects.Based on the stage division model of public opinion,this paper studies the development of public opinion about four Internet financial violations,and conducts text analysis experiments on the data in different stages to explore the law of development and change.Text analysis experiment includes topic extraction and emotional tendency analysis.Topic extract experiment is based on Text Rank and TF-IDF algorithms to extract keywords that can express the theme of text.In order to analyze the theme changes intuitively,this paper relies on tools to draw word cloud for the sake of getting theme changes in different stages of public opinion development.Emotional inclination analysis uses the Bert model proposed by Google,and uses Pytorch to realize it.From the perspective of public opinion development,most events fit the three-stagemodel development law of "Incubation period-Diffusion period-Recession period".In the mean time,the follow-up progress of events may cause netizens to pay attention again,following with multiple outbreaks.As for themes of public opinion,the experimental result shows that there is obvious difference in different periods.The result of sentiment analysis shows that the sentiment of public opinion is always negative in each development period.Finally,according to the law and influencing factors,this study puts forward suggestions to control the development of public opinion.This study analyzes the public opinion text of Internet financial violations,mining theme and emotional polarity from massive text flow,and studies its rules to provide the basis for public opinion prediction,which has innovative and reference significance.