Intelligent Supervision Of Network Public Opinion Early Warning Risk Based On Graph Neural Network

With the rapid development of the Internet,more and more people are freely and freely sharing their views and opinions on online platforms.The impact of online public opinion on political life order and social stability is increasingly deepening.The use of computer technology to identify public opinion information from a vast amount of information for analysis and provide intelligent decision-making to achieve intelligent supervision has important social significance.However,there are still the following problems in the intelligent supervision of online public opinion risks:the complexity and variability of online public opinion risks are difficult to comprehensively evaluate;There are significant differences in the structure of online public opinion information and strong sensitive characteristics;The manual disposal plan is difficult to ensure timely and accurate management of public opinion.In response to the above issues,this article mainly conducts the following research work:In response to the complex and variable risks of online public opinion,which are difficult to comprehensively evaluate,this paper studies and designs a risk assessment index system for online public opinion warning.Based on the characteristics of online public opinion information,the risk of online public opinion is evaluated from the degree,type,and scope of harm;This paper designs and implements a risk assessment model for online public opinion warning.Based on grammatical relationships,the model identifies online public opinion risk information based on sensitive and emotional words.The accuracy of this model in online public opinion detection is 3.89%higher than that of the comparative model.In response to the problem of significant differences in the structure of network public opinion information and strong sensitive features,this paper studies and implements a public opinion risk classification model based on heterogeneous graph neural networks.By utilizing the characteristics of graph structure and combining the advantages of graph neural networks and pre trained models,the model’s ability to capture sensitive features is increased,and the structural differences between texts are alleviated.The results of comparative experiments and ablation experiments indicate that the accuracy of this model in sensitive text classification tasks is improved by 3.52%compared to the best performing baseline model.In view of the problem that it is difficult to guarantee timely and accurate management of public opinion by manual disposal scheme,this paper studies and realizes the automatic generative model of online public opinion risk disposal scheme,builds a decision database based on historical public opinion governance cases,and generates a disposal scheme by integrating various evaluation results.The experimental results show that the sentiment classification module has improved by 2.48%compared to the comparison model,and this model can fully utilize the content and additional information of public opinion texts to generate disposal plans.Finally,this article designed and implemented an intelligent monitoring system for online public opinion warning risks,and its usability was proven through functional testing.Public opinion management personnel can monitor network public opinion risk information through visual systems,and timely manage network public opinion based on system evaluation results and intelligent decision-making suggestions,maintaining the security and stability of the cyberspace.