Research On Opinion Dynamics In Microblogging Network Based On Heterogeneous Relationship

With the rapid development of Web 2.0 technology and mobile communication in recent years, microblog has gained widely attention, and it has become important media to communicate and express for the public. Meanwhile, microblog has been a double-edged sword, due to its nature characteristics, such as interactive, real-time and anonymity, and it makes a profound impact on national security and social development. Therefore, the research on opinion spread modeling in microblog is quite valuable from both theoretical and practical perspective.From the viewpoints of data-driven and model-driven, this dissertation mainly focuses on the basic properties of microblogging network and opinion spreading models in microblogging network. In detail, we highlight the main contributions of this dissertation as follows.1. For the basic properties of microblogging network, we study on measurement and analysis of microblogging network in China. For China’s largest microblogging system- Sina Weibo, we collect relevant data by crawler technology, and measure the topological characteristics and user behavioral characteristics of Sina Weibo. Compared with other foreign online social networks, we find that Sina Weibo has the same topological characteristics as others, such as small world, scale free, assortative mixing. However, there exists some differences in the characteristics of degree distribution, correlation of indegree and outdegree, due to the distinctness of system design mechanism, and Chinese- foreign cultures. Moreover, microblogging user behavior exhibits typical day-cycle and week-cycle patterns.2. For the characteristics of implicit user relationship in microblog, we study the mining method based on rules. We divide the implicit relationship mining problem into two steps: first one is the map from text content to entity relationship semantic model, and the second one is the map from entity relationship semantic model to emotional value. Based on the structural characteristics of miroblogging text, we propose some basic algorithms to automatically extract and quantify implicit user relationship from microblogging text, such as miroblogging ananphora resolution algorithm, miroblogging entity semantic relationship extraction algorithm, and miroblogging entity semantic relationship with emotional computing algorithm. Finally, we build the entity relationship with positive and negative attributes in miroblogging network.3. For the characteristics of heterogeneous relationship in microblog, we study constructing and analyzing methods of a social network for users discussing a topic. By aggregating the explicit user behavior relationship and implicit entity relationship, we propose a method to calculate the relationship strength, and build the social network with heterogeneous relationships in mircoblog. Finally, based on the microblogging dataset of “Fang Zhouzi VS Han Han”, we study the structural properties of the empirical social network, and find the core entities in this topic by k-core decomposition.4. For signed social networks, we study the dynamical modeling method of discrete opinion spreads in a social system. We extend the basic SIR model to simulate the process of discrete opinion propagate in signed networks, and build the general SIR model. Then, we analyze the dynamical behavior of the model, and propose a method to compute the opinion spreading threshold in signed networks with heterogeneous relationships. Finally, by comparative analysis of model evolution in microblogging signed network and empirical opinion spreading process, we verify the validity of the model.5. For social networks with heterogeneous relationships, we study the dynamical modeling method of continuous opinion spreads in a social system. Using continuous opinion framework, we propose a novel opinion dynamics model, which is based on the classical opinion dynamics models. Then, we analyze the stability of the model for the simple triangular structure network, and verify the results by simulation results. Afterwards, we analyze the evolutionary phenomena though opinion dynamics in some typical topologies of complex networks, and find that the main factors affecting the final public opinion are average degree of networks and social harmoniousness. Finally, compared with the empirical data of opinion spreading in microblog, the simulation results are found that the model can be better explain the public opinion evaluation in microblogging network.In summary, we aim at modeling and analysis of public opinion spreading in microblogging network, combining the empirical data with theoretical models. Some key techniques, including measurement and analysis of microblogging network, construction and analysis of networks with heterogeneous relationships, opinion spreading dynamics in signed networks, opinion dynamics in networks with signed and weighted relationships, are intensively studied. These techniques have brilliant perspective on the formation, analysis and supervision of public opinion in microblogging network.