Research On Algorithms Of Network Based Opinion Propagation

Technology develops constantly in today's society.Internet affects the political, economic and cultural life increasingly which draws great attention of the whole society. Report to the Seventeenth National Congress of the Communist Party of China on Oct. 15,2007 points out this problem that "We will strengthen efforts to develop and manage Internet culture and foster a good cyber environment." The emergencies of public opinion in the Internet have the characteristics like larger scale,faster diffusion velocity and more affected people after the emergencies has transferred from traditional ways to Internet.The traditional method in communications field can just give a description of the phenomenon and quantifiable research cannot be done.The continually emergence of public opinion events urge the deeply research of network based opinion dynamics. Therefore,both the analysis of network based opinion dynamics and the theoretically explanation of formation and propagation have significant meanings.The thesis studies the theories of network based opinion dynamics from the view of evolution process of virtual community networks,network structures and user behavior analysis based hot topic detection,personal strategy selection in opinion evolution,group size in opinion formation and the influence of intervention strategy on opinion evolution.Some schemes are provided in this thesis to resolve the issues.The research work of this thesis is supported by the Cultivation Fund of the Key Scientific and Technical Innovation Project under Grant 707006,and special foundation from Key Laboratory of Communication & Information Systems(Beijing Jiaotong University) and Beijing Municipal Commission of Education..The main innovations of the thesis are as follows:1.In the field of propagation media,we investigate and simulate the evolution process and topological features of the virtual community networks.We founded that the joint of nodes,joint of edges,degree distribution of networks, average degree and the correlation between the degree of nodes and its time in the network are different from the scale-free networks model.With statistics and analysis of some of the characteristics of an actual Internet forum,we propose a new network evolution algorithm,namely the virtual community networks construction algorithm.The simulation results show that the networks generated from our algorithm has the same characteristics of real virtual community networks.2.In the field of hot topic detection,a network structures and user behavior analysis based novel method to detect the hot topics mainly focused on the Internet forum was proposed.The solution considers two important characteristics of Internet forum:the network structures and the user behavior patterns.First,we construct a network of the users' "replied to" relationships in a given Internet forum.Second,behavior patterns for all users are computed. Finally,we combine the results from first two steps to determine which topics are hot topics.Results of our empirical experiments show that our approach can identify hot topics easily and effectively.The proposed solution can enhance the controversial and sensitive topic detection.3.In the field of individual's interaction,a model was established for opinion evolution,simulations and analysis were carried out.First,the model allowing people to change their strategies when interacting with others in opinion evolution were established based on human behavioral patterns in real life.A model of people's memory in opinion evolution was made.Second,the model with different sets of parameters was simulated for many times and the results made a good agreement with what happened in real life.Justification conclusion was achieved for personal strategy selection in opinion evolution.4.In the field of interacting groups,a model from the social psychology points of view was proposed.The model considers both parts(positive opinions and negative opinions) of some connected people.We simulated the formation process using the proposed model,results show that the model is useful for explaining how group/herd mentality causes prevalence and how conformity lead to balance.5.In the field of public opinion intervention strategies,models and intervention strategies were established for opinion evolution,simulations and analysis were carried out.First,one dimension and two dimension models were established based on human behavioral patterns in real life,results of these two models made a good agreement with opinion evolution in real life and aggregation phenomenon emerged in the simulations.Second,two models of intervention strategies were established based on the former opinion evolution models, simulations and analysis were done for these two models.Simulation results showed that different intervention strategies have different kinds of influence on the ultimate results of public opinion evolutions and could be used in different cases.The research work on network based opinion dynamics will be further on with the extension of its application.