Research On Network Public Opinion Early Warning Of Public Emergency Based On GA-BP Neural Network

With the continuous progress of science and technology,the role of the Internet in the process of social operation and governance is becoming more and more obvious.In this context,the public began to pay more and more attention to public emergencies and actively participate in the discussion.Hot topics frequently appeared in the public opinion field,the public self-expression became more and more casual,and wantonly spread negative emotions,and the differences of groups gradually appeared.In recent years,public emergencies have occurred from time to time.Once offline events spread on the Internet,they will cause heated discussion among netizens.At this time,if the government lacks the awareness of timely and effective response,it will accelerate the spread of public emergencies in cyberspace,which is easy to aggravate the dissatisfaction of netizens,resulting in the crisis of online public opinion,Therefore,it is of certain value to study the early warning of network public opinion for public emergencies.This thesis proposes to establish a network public opinion early warning model for public emergencies,hoping to enrich the research in the field of network public opinion early warning,and provide reference for public opinion monitoring and governance of relevant departments.This thesis combs the research status of network public opinion of public emergencies,expounds the relevant theoretical research,selects 20 public emergencies with great impact since 2015,crawls the data from the event incubation period to the extinction period on the microblog platform as the research object,and analyzes the impact on network public opinion of public emergencies from the public opinion subject dimension,event dimension and media dimension in combination with the views of previous studies,As the primary indicator of the early warning index system,the main dimension includes four secondary indicators: official response,response channel,emotional tendency of netizens and public opinion heat.The secondary indicators of the media dimension include in-depth media reports and negative reports.The secondary indicators of the media dimension include media in-depth reports and negative reports.The influencing factors are quantified to build the network public opinion early warning system of public emergencies in this thesis.Then,grey correlation analysis and K-means method are used to comprehensively evaluate public opinion events,and determine the early warning level of public opinion events.Finally,public opinion events are divided into four levels: extreme danger level,danger level,warning level and slight level,which are used as the training samples of early warning model.Taking public emergencies as an example,considering the strong self-learning and self-adaptive ability of BP neural network,this thesis can deeply dig and effectively use the complicated historical data of public opinion in public emergencies network,make up for the shortcomings of BP neural network through GA genetic algorithm,and finally establish the public opinion early warning model of public emergencies network based on GA-BP neural network to solve the problem of strong subjectivity of the existing early warning model,It provides an effective and feasible method for reasonably and accurately predicting the network public opinion of public emergencies.Finally,through empirical analysis,the accuracy of the model is 88.5%,and the model performance is good,which shows that the index system and model constructed in this thesis are feasible and can provide reference for government public opinion governance.