Research On Early Identification And Evolution Model Of Social Network Public Opinion Based On Deep Learnin

With the continuous iteration and development of Internet smart application devices and network technology,public opinion sentiments can be published in real time on online social platforms.The ensuing false information and negative public opinions have caused harm to social public safety and the online environment.Timely and accurate identification of false information in online public opinion and analysis and prediction of the propagation mechanism and evolution law of public opinion are effective means to eliminate the harm of public opinion.However,information at the early stage of public opinion dissemination is often sparse,making it difficult to accurately identify its authenticity.The chaos of complex networks also makes it difficult to generalize and analyze the propagation mechanism and evolutionary laws of public opinion networks.Therefore,it is important to establish an effective model for discerning the authenticity of public opinion in the early stage of its dissemination,to study the mechanism of public opinion dissemination and analyse the trend of public opinion dissemination,and to propose methods for controlling public opinion.It is an important way to enhance the ability of comprehensive social governance.This study models and analyses online public opinion.It achieves the detection of false information at the early stage of public opinion dissemination.On this basis,the mechanism of public opinion dissemination is studied,an evolution model of public opinion is established,and the threshold of public opinion dissemination is searched for to determine the dissemination window.Effective control methods are proposed for different stages of public opinion development.The main contents of the research are as follows.(1)In order to obtain more data features at the early stage of opinion propagation.Considering the small-world property of social networks,this study proposes a rumor detection method based on graph attention networks to fully learn the structure of rumor propagation and the deep representation of text content.Early rumor detection is performed by contextualizing the text with semantic association features and combining the full content of the source text.In particular,the proposed method has higher accuracy and better robustness in the context of sparse data in the task of early opinion truth detection.(2)In this study,a definition of circle-stratification networks segmentation is proposed,and the different stratifications obtained by segmentation help to discover and capture key nodes in the network.Afterwards,the propagation dynamics of public opinion in dynamic circle-stratification networks are mechanistically analyzed and a dynamic circle-stratification network-SEIR public model is constructed.The thinking time lag parameters in the model are used to explore the window period of opinion propagation,quantitative expressions for the control parameters of each stratification are given according to the set of control parameters,an pinning control methods is developed.The experimental results show that the proposed stratification division method can help solve the curse of dimensionality and reduce the scale space of public opinion propagation.The pinning control of the control parameter set can achieve an asymptotically stable state within the minimum propagation threshold within the system.(3)Based on the proposed model and method,this study conducted an empirical analysis on a typical dataset of microblogs related to the COVID-19.The analysis results demonstrate that the method proposed in this study can not only achieve highprecision authenticity discrimination at the early stage of public opinion dissemination,but also verify the rapidity and stability of the model in controlling the implication of different stratifications.The final findings of the study provide reference suggestions for emergency planning decisions in response to the spread of negative public opinion.