Research On Evolutionary Dynamic Model Of Public Opinion Based On Scale-free Network

In the era of the new media and network supervision with the rapid development ofInternet, network public opinion is becoming an important medium of impacting themainstream of social orientation. The research of public opinion has gradually extendedfrom the communication field to interdisciplinary fields including artificial intelligence,complex system, numerical calculation, physics, cybernetics and so on. However, thecurrent study is limited to the construction of the various models and thus cannot fullyreflect the evolution of public opinion in real life. Also the establishment of interactionrules in those models has not considered the interpersonal networks, so there hasn t apositive and effective way to guide the development of public opinion. Therefore, thestudy on the evolutionary dynamics models of public opinion becomes one researchhotspot in recent years.Based on the comprehensive overview of the evolutionary dynamics models ofpublic opinion at present, difficulties and deficiencies existed in the research of publicopinion are analyzed in this thesis. For this case, from the perspective of complexnetworks and emergent computation, the formation of public opinion as well as itsinternal and external factors is first deeply analyzed in this thesis; then the study focuseson the models of public opinion. The main work in this thesis is organized as follows:Firstly, the existing evolutionary dynamics models of public opinion can t fullyreflect the law of public opinion in real world. For this problem an improved KH(Krause-Hegselmann) Model based on scale-free network is proposed. According to thescale-free property of the Internet, a scale-free network is built as the initial network ofinterpersonal relationships. And the KH evolutionary dynamics model of public opinion,a kind of bounded confidence models, is selected and analyzed. Combined with thepropagation law of public opinion in the real world, the environmental factor andpersonality of agents as well as intimacy are considered in the improved model. Theresults of simulations and comparison experiments show that formations of publicopinion presented by the improved model are much richer than the KH model. Theimpacts of environmental factors and intimacy on the formation of public opinion aremore realistic.Secondly, for the problem that few researchers verify the correctness of theirmodels using real data, the real data from Internet using the web crawler is acquired in this thesis. Then the real data is imported into the improved model for simulationanalysis after manual handling. Experiment results show that the improved model wellmeets the law of the development of public opinion in realistic society, which hascertain rationality and validity.Thirdly, for the deficiencies in the research of the evolutionary dynamics of publicopinion under the perspective of the emergent computation, trends of network topologyduring the evolution of public opinion based on the BA (Barabasi-Albert) network areanalyzed in this thesis. Simulation results show that the topology, the network clusteringcoefficient and the degree distribution change in a regular way during the evolution.Furthermore, the emergent behavior of this model is quantitatively assessed bycalculating the power spectral entropy and other indicators of dynamics characteristics.And it can be summed up that those indicators show regular changes, which provides atheoretical basis for the guidance and control of the public opinion. Hence the abovestudy has certain practical significance.Finally, in Matlab2010b platform, the prototype system about the dynamicevolution of network public opinion is designed and implemented. It contains all of theabove mentioned functions and provides a foundation for the further research of theemergent behavior in network public opinion.