Nuclear Public Opinion Sentiment Analysis And Early Warning Methods

With the development of internet information technology,many people have participated in online social networking,which has made the status of online media continue to rise,and at the same time has accelerated the scope and influence of online public opinion.The early warning of public opinion is of great practical significance for the government to guide public opinion.The timely detection of negative public opinion information is conducive to the government's correct decision-making.Therefore,how to effectively analyze and deal with Internet public opinion events has become a research hotspot.Although the current public opinion detection system has achieved certain theoretical and practical results,there are still many problems that need to be solved,such as the incompleteness of the structure of the public opinion index system,the time of the public opinion explosion,and the incompleteness of the domain feature lexicon.This article mainly focuses on the monitoring of public opinion in the nuclear field,and mainly includes three aspects: the discovery of new words in the network of characteristic thesaurus in the nuclear domain,the analysis of the sentiment orientation of nuclear microblogging,and the early warning ofnuclear public opinion.The discovery of new words in the nuclear domain feature thesaurus network is mainly constructed by calculating the similarity of nuclear-related words.This paper uses information entropy and the synonym forest+word2vec to calculate the similarity of nuclear words.Firstly,the key words are extracted by the information entropy technology,then the word similarity calculation method is used to calculate the word similarity between the words synonym forest +word2vec,and the similarity words satisfying a certain threshold value are selected and finally passed through the experiment.The effectiveness of the improved algorithm is demonstrated.The analysis of the emotional tendency of nuclear-related microblogs is based on the combination of RAE+Dropout to determine the emotional bias of microblog.In this paper,using the RAE model can transfer text features,combined with the advantages of Dropout to improve the generalization of the model,an improved algorithm is proposed.This algorithm improves the generalization ability of the sentiment analysis model,and finally validates the improved model through experiments.Effectiveness.The early warning of nuclear public opinion mainly builds models by extracting early warning indicators for public opinion,and then sets thresholds for early warning.This article analyzes the characteristics ofWeibo texts,extracts the index characteristics of the nuclear weather warning,and uses the posting time,the amount of forwarding,the amount of comments,and the amount of praise as the index system of the nuclear weather warning,and then builds a model for the warning index.The rate of change of negative information is used as the threshold for early warning of nuclear public opinion,and early warning is conducted through the size of the threshold to control the impact of nuclear public opinion within a controllable range while minimizing the harm of nuclear public opinion.