Research On The Grading And Early Warning Model Of Network Public Opinion Under The Background Of Big Data

In the era of the network information open to the media,everyone can freely express views,opinions,and emotions in an informational way.In addition,the Internet has the characteristics of wide audience,fast transmission speed,transcending the space and time,etc.,as well as the limitations of the major media platforms on the number of words and the length of videos edited,so that the accuracy of relevant information transmission is not high,which makes it easier to transform the network public opinion information into the network public opinion crisis information.Compared with the traditional public opinion crisis,the public opinion crisis evolved from online public opinion has a high probability of outbreak,a short duration,a larger scope of influence,and the government control has become more difficult.Therefore,it is necessary and urgent to study the early warning of online public opinion crisis.The most effective way to carry out early warning of online public opinion is to be able to identify online public opinion crisis information in the early stage of the emergence of crisis information,and then take corresponding measures to implement control.In the context of big data,this paper analyzes the research status and development trend of domestic and foreign network public opinion early warning.By using the methods of literature analysis,comparative analysis and the popular data mining algorithm,this paper studies the index system,classification criterion and analysis method of network public opinion crisis early warning from multiple levels and perspectives,aiming to build Intelligent integrated network public opinion early warning automatic identification system.Based on the inadequacy of research on online public opinion today,this paper discusses the characteristics and evolution law of online public opinion from various aspects.On the basis of following the principles of feasibility,comprehensiveness,scientificity,and quantitative as the main,qualitative as the auxiliary,we first constructed an Three-level,four-category,13 second-level indicators of network public opinion early warning indicator system;Next,driven by data,the gray correlation analysis method and clustering algorithm are used to implement the public opinion event grading process.The gray correlation analysis method introduces the idea of the entropy weight method to determine the comprehensive evaluation index of public opinion event heat,and the clustering algorithm is used to determine the classification threshold of public opinion events.Finally we divide the public opinion events into three categories: mild,warning and serious;Then,the random forest-sequence backward selection algorithm is used to extract the important public opinion early warning feature indexes in the construction of the early warning model,and the improved stack combination algorithm is used to construct the network public opinion early warning recognition model;Finally,the model is verified by using the data of the network public opinion case database,and compared with other models,the rationality,availability and effectiveness of the improved stacking model are proved.