Evolution And Prediction Model Of Network Public Opinion Heat Based On Text Mining

The popularity of public opinions on social networks is a direct indicator to measure the degree of attention paid to public emergencies.The vast number of netizens publish and share information through online social networks to express their views and positions on hot social events,which promotes the dissemination of information and the evolution of public opinion.The heat reflects the development state of network public opinion from germination to expansion.The communication of network public opinion is profoundly changing and reshaping the social public opinion ecology,and has a direct impact on the thoughts and behaviors of the vast number of netizens,which leads to public security incidents and even the formation of public opinion crisis.Today,as the scale of social network users continues to increase,it is of great practical significance to study the evolution rules of public opinion in different communication periods and explore heat prediction methods to improve the prediction ability of online public opinion,for creating a good ideological and public opinion atmosphere and maintaining social order and stability.However,in the face of complex network public opinion environment,distance to the heat of the existing research methods applied in the management of public opinion,there are some problems,which includes both a vast amount of network data is converted to heat,quantitative data,including the existing heat evolution prediction method of social network public opinion characteristic of the evolution and the one-sidedness and limitations of multiple factors to consider.This thesis revolves the public opinion of representation and quantification of heat ","public opinion heat evolution" and "the multi-factor coupling heat prediction" three key issues,from the public opinion of the heat index based on the interaction and emotional conflicts building,heat network public opinion evolution research based on text mining,neural network based on the depth of multi-factor coupling heat to predict three aspects to carry out the research,The main work contents are as follows:Based on the interactive behavior and emotional state of netizens in the process of public opinion transmission,the heat quantization algorithm of network public opinion is constructed based on the characteristics of interactive behavior and emotional conflict.Firstly,based on the interactive behavior characteristics of netizens,information gain rate is introduced as the weight index to judge each interactive attribute,and the interaction heat caused by complex interactive behaviors in the evolution of public opinion events is measured.Secondly,the emotional heat of public opinion is reflected from the emotional dimension of netizens,and the convolutional neural network and other methods are used to mine emotions from the rich text resources of public opinion,and the variance of netizens’ emotional tendency is used to define the emotional dynamic intensity of public opinion,and then the potential emotional heat situation of public opinion is measured.Finally,the potential communication heat of social events is identified through the representation of public opinion communication through two aspects of interactive behavior heat and emotional dynamic heat.Aiming at the complex online social network public opinion events,the heat measurement algorithm of public opinion based on interactive behavior and emotional conflict is introduced to study the evolution law of online public opinion heat.Firstly,based on the changing characteristics of user interaction behavior,the heat evolution stage of public opinion communication is divided,and the LDA theme model is used to mine the hidden topics of the topic,so as to provide a basis for relevant departments to timely discover the potential trend of public opinion.Then it analyzes the cross-evolution law of interactive behavior and emotional heat of public opinion in different evolution stages.The results of evolution analysis show that the quantification model of communication heat constructed in this thesis can effectively utilize the effective emotional influence power that causes the intensification of public opinion in the process of communication,and provide important support for the accurate identification of the evolution law of public opinion and the future trend of heat.In view of the decomposition and prediction of the driving factors of public opinion heat based on feature engineering,a multi-feature coupling heat prediction model based on deep neural network was proposed based on the formation mechanism and characteristics of microblog public opinion.The model extracts each characteristic index from the complex public opinion data and makes grouping according to the characteristic attribute.The user influence and sequence characteristics of interactive network structure were extracted by short and long time memory network,and the public opinion text structure,sequence characteristics and historical public opinion heat were fused by fully connected network to predict the state of public opinion heat.The experiment shows that the multi-feature coupled public opinion heat prediction model based on deep neural network can take into account a variety of factors affecting the heat evolution,fully mine the sequence features of text content and interactive network structure,and the heat prediction algorithm has certain advantages in the prediction ability compared with the existing heat prediction algorithm which has achieved better results.