Research On The Evolution And Early Warning Model Of Internet Public Opinion

Internet public opinion is increasingly becoming the main factor affecting the Internet and social stability,especially in recent years,Internet public opinion crises have been on the rise.Compared with the traditional public opinion crisis,the Internet public opinion crisis has a higher probability of occurrence,faster risk transmission,and a wider scope,and increased difficulty in management and control.Therefore,it is urgent to improve the early warning ability of Internet public opinion.The most effective method is to identify Internet public opinion crisis in the early stage of its occurrence and implement corresponding management and control measures.This paper first describes the current situation of Internet development and public opinion in China,and summarizes the domestic and foreign research status of Internet public opinion early warning and topic evolution model.Secondly,defining Internet public opinion,Internet public opinion crisis,and Internet public opinion early warning,Back Propagation Neural Network and OLDA,focuses on comparing the methods of building Internet public opinion early warning model,and expounds the advantages of Back Propagation Neural Network.The focus of this paper is to build a Internet public opinion early warning model.First of all,according to the construction principle of the indicator system,combined with the characteristics and evolution rules of Internet public opinion,the primary indicator system of Internet public opinion early warning was built.Then,driven by data,the public opinion indicator data related to the “3.21 East Airline Crash” event was collected from Sina Weibo and Baidu and preprocessed.Finally,4 main indicators,of which 10 secondary indicators constitute the Internet public opinion risk early warning indicator system.Secondly,use the gray correlation analysis method and KMeans clustering algorithm to achieve the Internet public opinion early warning level classification,and divide the events in different time slices into three levels: slight,warning and serious.Then,we use the OLDA to establish a public opinion theme evolution model,and analyze the theme changes of Internet public opinion events with different warning levels.Finally,build a Internet public opinion early warning model based on Back Propagation Neural Network.Through the analysis of the early warning results of the test samples,the early warning accuracy of the model reaches 90.91%.Therefore,in the actual early warning application,the Back Propagation Neural Network can be used to warn the Internet public opinion events.