【作者】 张伟；

【导师】 何明升；

【作者基本信息】 哈尔滨工业大学 ， 管理科学与工程， 2014， 博士

【摘要】 随着网络社会的崛起，作为现实社会“晴雨表”的网络舆情逐渐成为一种不可忽视的频发性社会现象。通过聚合民意，网络舆情能够形成强大的社会合力，并产生深刻的社会影响。在此背景下，研究网络舆情的发生发展规律，对网络舆情演化的动态过程进行系统分析，具有重要的现实意义。本文在梳理国内外相关研究成果的基础上，综合运用复杂网络分析、理论建模与计算机仿真等研究方法，通过对网络舆情演化过程中的个体行为和个体交互特征进行合理抽象，构建了复杂社会网络中的网络舆情演化模型，并对模型进行了深入的仿真分析。具体的研究工作主要从以下几方面展开并得出结论：首先，根据社会动力学模型的建模范式，分析网络舆情演化模型三个方面的要素构成，提出网络舆情演化的交融-分阶结构。网络舆情演化是一个多主体协同的过程，以具有小世界性、幂律分布、异配性和富人俱乐部效应等拓扑特性的复杂社会网络相互连接的多样化行为主体借助弱连接的桥接作用、中心节点的辐射效应以及层级化的交互过程以自组织协同的方式推动了网络舆情的演进。网络舆情演化不是一个简单的单向线性过程，它具有复杂性和动态性，其中交织着网络空间中由各行为主体之间的交互所引发的网络舆情观点聚合和网络舆情信息扩散这样两个分支环节，它们相互融合、相互作用，以一种混合连接的形态形塑了网络舆情演化的结构秩序。这一部分的内容为网络舆情演化模型的构建奠定了理论基础。其次，基于复杂社会网络理论和观点动力学模型的建模思想，构建了网络舆情演化中的观点聚合模型。在网络空间中，以一定网络结构连接起来的多样化行为主体围绕某一舆情事件形成的观点经历了紊乱无序到一致或相对一致的动态演进过程；从微观的个体层面来看，这一过程也是网络空间中的多样化行为主体按照一定规则不断进行观点更新的动力学过程。这些规则可以归结为个体观点在有限空间内连续取值、个体遵循有限信任原则从邻域个体中选择交互对象并按其对他人观点的接受度进行观点更新、个体连接而成的网络随着人员的流程而不断变化等三个方面。利用计算机仿真的方法，本文对建立的网络舆情观点聚合模型进行了仿真分析，揭示了个体观点接受度、信任阈值、个体观点初始分布、网络结构及意见领袖对网络舆情观点聚合的影响。再次，通过分析网络舆情信息扩散中的个体作用模式，提出了基于复杂社会网络的个体接触过程，而网络舆情信息正是藉由个体接触实现了在整体网中的扩散。参照仓室模型中SEIR模型的个体状态划分方法，本文将网络舆情信息扩散中的个体划分为非知情、知情、传播和移出四种状态，并且按照现实中的网络舆情信息扩散状况界定了各状态之间的转移过程和转移概率，进而构建了基于复杂社会网络的网络舆情信息扩散模型。通过计算机仿真分析，本文揭示了网络舆情信息扩散效果对模型主要参数、网络结构特性及初始传播者数量和所处网络位置的敏感性。最后，遵循高影响力和强时效性两个原则选取“1122青岛输油管道爆炸事件”作为样本案例，对网络舆情观点聚合和信息过程进行了案例分析。通过对案例的系统分析，本文呈现了特定网络结构下受个体观点初始分布和意见领袖影响的网络舆情观点聚合过程，分析了网络舆情信息扩散效果对网络结构和初始传播节点的敏感性，进而为前文建立的网络舆情演化模型进行了验证。更多还原

【Abstract】 Network public opinions are like prisms that well refracting rational social relationship in real life. With the rising of network society, network public opinions, the barometer of real society, is becoming a frequent social phenomenon that can’t be neglected. It can form significant scale effect and further generate deep-going social effect with gathering popular will. In this case, the study of its developing regular pattern and systematic analysis of its dynamic process have important reality meaning for ensuring social balance, preserving social stability and promoting social harmony. On the base of systematic sorting and summarizing related research results in and abroad, this paper apply means of complex network analysis, theoretical modeling and computer emulation comprehensively to abstract reasonably individual behaviors and interactive characteristics in the dynamic procedure. In this way, the dynamics model of network public opinions in complex social network is built and deep emulation analysis of the model is preceded. This study is based on the view of complex social network to recognize inner rule of network public opinions’ dynamics, and is mainly developed and getting result through the following aspects.Firstly, this paper defines the connotation of network public opinions’ dynamics through sorting related research results and presents the blend-sublevel structure of dynamics. It points out that the dynamics is not a simple one-way linear proceed and it’s complex and dynamic, instead. Within it there are two branch links as opinion gathering and message diffusion which are caused by behavioral agent’s interaction in the network space. The two links mix and affect mutually, building the dynamics’ structural order in the form of mix joint. In the view of social dynamic’s research paradigm, the macroscopic level as the rising of complex group and social phenomenon has great relationship with the microscopic level as the simple individual behavior. For this reason, this paper further analyses the basic structure and topological property of dynamics space, proposing the multi-agent interaction model and holding the view that the diverse behavior agent connect each other in the way of complex social network. The agents use the bridging effect of weakly-coupled, radiation effect of center point and interactive proceed of stratification to push the dynamics’ developing in the form of self-organized cooperation. The above analysis has constructed the base of dynamics model.Secondly, based on the theory of complex social network and modeling view of standpoint dynamic model, the essay abstracts gathering rule of individual opinions in the dynamic proceed and constructs opinion gathering model in complex social network. In the network space, opinions of diverse behavioral agent about a certain public sentiment have experienced from chaos to agreed or relatively agreed. From the microscopic view, this process is the dynamics proceed that the diverse behavioral agents constantly renew their opinions in some certain rules. These rules can be summed up as three aspects that individual opinions evaluate continuously in limited space, individuals choose interactive agent in neighborhood space following limited trust and renewed opinions on his extent of accepting others’ opinions and the network connected by individuals contently changes with the flow of people inside. This research uses the method of computer simulation to analyze the opinion gathering model. In this way, it announces the impact of acceptance degree of individual opinions, trust evaluation, initial distribution of individual opinions, and network structure and opinion leaders for the opinion gathering of network public opinions.Thirdly, through the analysis of individual affect model in the information diffusion, the paper proposed individual contact process based on complex social network. For individual contact, the network public opinions fulfilled diffusion in the whole network. Referring to the dividing method of SEIR model’s individual states in the Bin effect, the agents are sorted as unwitting, witting, spread and shift out. And as the same in the real information diffusion, the transform process and rate of the states are set to construct the information diffusion model. Using computer simulation analysis, the essay shows the stimulation of information diffusion effect to the main parameter, network structure feature, quantity of initial spreaders and location in the networks.Finally, following the rules of heavy impact and great timeliness, the essay chooses1122explosion of oil pipeline in Qingdao as the sample to analyze the opinion gathering and message proceed in it. Through the systematic analysis, this paper presents the opinion gathering proceed under certain network structure and influenced by individual opinions’ initial distribution and opinion leaders. In addition, analyzing the stimulation of the message diffusion effect for the network structure and initial spread point, and further testing and verifying the dynamics model of network public opinions.更多还原