Online Public Opinion Prediction Model Based On Microblog Sentiment Analysis

More and more people are willing to speak out for online public opinion events through online platforms,and they indirectly participate in social construction by expressing opinions and making suggestions on relevant public opinion events.Therefore,the sentiment analysis of microblog texts,and the calculation of the sentiment intensity of hot topics through the sentiment analysis results and the construction of a network public opinion prediction model have become an important means of public opinion analysis.This paper constructs a network public opinion prediction model based on Weibo sentiment analysis,and combines Weibo hot topics to classify the risk level of Weibo public opinion,judge the development trend of public opinion,and formulate a public opinion disposal strategy.The main research contents are as follows:(1)A microblog sentiment analysis model based on the combination of convolutional neural network and sentiment features is proposed.The attention mechanism layer is introduced between the word vector input layer and the convolution layer to reorganize the microblog word vector.The reorganized word vector is used as the input vector of the convolutional neural network,and the convolution operation is used to learn the context of non-consecutive words.The related semantic features between the microblog texts are extracted,and the semantic feature vectors of the microblog texts are extracted.The emotional features of emotional words,degree adverbs and transition words in the text are screened through the traversal rules,and the emotional information collection is obtained to build an emotional resource library and extract emotional feature vectors.The extracted semantic feature vector and sentiment feature vector are spliced into a new feature vector to train the model to solve the problem of contextual semantic incoherence and part-of-speech neglect.(2)A method for calculating the comprehensive emotional intensity of network public opinion is proposed.First,combined with the number of punctuation marks and emoticons in the microblog text containing modal particles,the emotional strength of the microblog text is calculated by setting the weight,so as to obtain the emotional strength of the microblog text.Secondly,the growth rate of the topic text and the increase of comments,likes,etc.within a specified period of time are combined with the calculation method of emotional strength to calculate the emotional strength of Weibo topics.Finally,the comprehensive emotional strength is obtained by calculating the emotional strength of the microblog text and the topic emotional strength of the microblog.(3)To calculate the rate of change of topic heat on the comprehensive sentiment calculation results of Weibo obtained in a specific time period,to obtain the change rules of topic heat and sentiment intensity in different time stamps,so as to judge the development trend of Weibo public opinion events.Combined with the public opinion prediction level index,the microblog topics are divided into risk levels,and then the microblog data is used to verify and analyze the model to verify the feasibility and effectiveness of the model.