Research On Public Opinion Warning For Sudden Public Events Based On Analytic Hierarchy Process And Temporal Neural Network

Strengthening online public opinion warning for sudden public events is of great significance for promoting social harmony and healthy living of the people.With the fast development of the internet,the internet has become the fourth major media to reflect public opinion,following newspapers,radio,and television.The dissemination of online public opinion has a unique characteristic of independent intersection,so it is rapidly fermented by multidimensional factors and has become the focus of public attention.The frequent occurrence of sudden public events and online public opinion events has attracted high attention from the government and relevant departments.Therefore,conducting realtime research and monitoring of online public opinion on sudden public events,establishing an effective online public opinion warning mechanism for sudden public events,timely grasping public opinion trends,actively warning,and actively guiding public opinion are effective means and key links to prevent public opinion crises in sudden public events.In view of this,this article proposes a public opinion warning method for sudden public events based on Analytic Hierarchy Process and Time Series Neural Network.The research content and achievements of this article include:1)Establish a network public opinion warning indicator system for sudden public events.Exploring the evolution of network public opinion in sudden public events,we delved into its characteristics and propagation mechanism.Additionally,we examined the factors that shape its evolution.A system of early warning indicators for sudden public events,based on the principle of indicator establishment,was created from both static and dynamic views of network public opinion,with the main body,object,and dissemination of public opinion as the primary indicators.The relevant properties and quantitative methods of early warning indicators were defined.Next,the entropy method and factor analysis method are used to calculate the weights of each secondary indicator,and the comprehensive weights of the secondary indicators are obtained based on the formula.Finally,the public opinion warning level of the sample event is calculated through comprehensive weights as the expected output of the sample for subsequent warning model training.2)Excavation of sudden public events.This chapter mainly introduces the methods for mining sudden public events in public opinion warning,which is one of the key issues in obtaining the data required for public opinion warning.The article provides a detailed introduction and implementation of data collection,data preprocessing,text feature encoding,and clustering in public event mining.An empirical analysis was conducted using food safety emergencies as a case,and good results were obtained.Theoretical and practical significance is greatly increased by this research,which furnishes dependable data and technical aid for public opinion caution.3)Constructing a network public opinion warning model for sudden public events.This chapter proposes an online real-time public opinion warning model based on temporal LSTNet.This model sets different time periods and time windows,combines the original public opinion data obtained earlier,and uses the weighted public opinion data formed by the hierarchical indicator system as input.Time-series LSTNet,employed by AHP-LSTNet,offers a precise examination of time series,allowing it to capture the accumulation,correlation,and dependency properties of public opinion data in the temporal domain,thus enhancing the effectiveness of real-time public opinion alerting.This research has both theoretical and practical worth,and can furnish dependable scientific basis and technical backing for actual public opinion surveillance and early warning.