Research On Early Warning Method Of Internet Public Opinion Oriented To Event Group

The Internet has become a dynamic window reflecting public opinion.More and more people obtain information through the Internet and express their opinions and emotions in cyberspace.The collection of opinions,emotions,and behavioral tendencies is called online public opinion.The objective and accurate judgment of the development trend and situation level of the Internet public opinion is of great significance to the correct decision-making of the government management department,and it is also a necessary condition for maintaining the harmonious and stable development of the society.The existing research on the public opinion of the Internet is usually to regard the events in the network as a single event,and there is little research on the multi-event network public opinion.However,due to the diversity of information in the network,and events are not constrained by the space-time distance when traveling in the network,there is a general correlation between events.When the topics,subjects,emotions and other factors of the network events are the same or similar The public will usually discuss together these events with commonalities.The group of events formed by these multiple events will have a greater impact than a single event and a greater degree of harm.Therefore,the event group The perspective of research on the public opinion of multi-event networks is of great significance.In summary,the research work of this article includes the following aspects:First,the Internet public opinion forecast.This article proposes a network event heat prediction method from the perspective of similar trends in similar events.This method first classifies network events into different categories on the semantic level,and then uses the EKSC algorithm to cluster the time series of different types of public opinion events at the temporal level.The EKSC algorithm first decomposes the original time series,and the decomposition includes many For the physical component temporal information,the low-frequency component cluster center is used as the initial clustering center of the high-frequency component to enhance the clustering effect.After clustering,the class model of each small class is obtained,and finally,the category and the sum of the events to be measured are obtained.The mean square error of each small class model and the most suitable class model are selected.The class model is used as the prediction template of the event.The predicted value of the event to be measured is obtained through translation and scaling,and the experiment shows the effectiveness of the prediction method..Second,the research on early warning methods of Internet public opinion oriented to event groups.For the first time,from the perspective of event group,we made judgments on the situation of network public opinion development and the level of early-warning.First,we described the related elements of the event group and gave formulas for calculating the degree of correlation between events.Second,we have established an early warning index system for public opinion in the Internet.Based on the above,we established an early warning indicator system for the network public opinion of the event group,and gave a quantitative method and standardized method for the final indicator;then,we used intuitionistic fuzzy reasoning to judge the development situation and warning level of the Internet public opinion,and made intuition on the input variables.All factors deal with fuzzy measure and establish inference rules,reasoning to get the warning level,and use the change of the situation index in unit time as the deterioration rate of the event,analyze the development of the public opinion event through the rate of deterioration,and finally,use the method proposed in this article in 18 event groups.Experiments on group datasets show that the method has a good effect on the evaluation of network public opinion situation and rank judgment.