【作者】 朱国东；

【导师】 何德全； 刘云；

【作者基本信息】 北京交通大学 ， 通信与信息系统， 2009， 博士

【摘要】 互联网的飞速发展使之逐渐成为社会舆论传播的新势力。网络舆情能够在很大程度上反映宏观舆情的真实状况,因而受到从政府到企业的广泛关注。近年来,网络舆论对现实社会已经和即将产生什么样的影响,如何实现有效的网络舆论预测及引导等问题正在成为社会科学领域和自然科学领域中的研究热点。然而,已有的模型虽然能够很好地解释舆论演进中存在的一些现象,但大多没有考虑网络结构的影响,也不具有预测演进趋势的能力。本论文结合当前对舆论演进行为、预测技术和引导机制等的需求,对网络舆论的相关问题进行初步探索,主要内容包括:网络结构特性对话题传播和观点交换的影响、网络舆论演进的特征以及网络舆论演进趋势的预测等。论文的研究工作得到了高等学校科技创新工程重大项目培育资金项目(No.707006)“网络舆情传播与预警关键技术研究”的支持。论文的主要创新点如下:①研究了网络拓扑结构对话题传播过程的影响,重点考察了网络的全连通性、小世界效应和无标度特性对话题的传播速度、群体观点的稳定性的影响,以及初始就持有观点的个体的数量和分布特征与传播过程的关系,得出一些对舆论引导有参考意义的结论,如当网络平均节点度较大时,节点的异质性结构对话题的传播有明显的阻碍作用;增大网络的簇系数可以在一定程度上加快话题的传播速度;影响话题传播速度的网络拓扑特性间存在制约关系等。②研究了舆论演化过程所处的网络结构特性对观点交换过程的影响,给出了表征个体观点分散程度的参量的定义和描述群体观点变化趋势的参量的定义。得到的主要结论包括:网络的小世界效应和同质结构能够使个体间产生充分的信息交换,有利于观点的统一;网络的无标度特征会使观点的聚合过程加长,一般情况下持非大众观点的个体会始终存在;对于异质结构的网络,较高的分簇水平可以加速观点的交换过程,缩短观点的稳定时间等等。③对网络舆论演进过程的一些特征进行了分析,发现话题不同发展阶段呈现出某种意义上的同质性。如果将这种特征看作演进系统的惯性,惯性的存在将是对舆论演进趋势进行预测的重要基础;热点话题阅读次数与发布时间之间具有相关性,这关系到序列整体的发展趋势和对舆论干预方式的选择。④对网络舆论演进趋势的预测进行了初步的探索。提出并研究了基于时间序列理论的网络舆论发展趋势预测方法。数值分析和实证研究表明,使用时间序列的分析方法,能够对舆论演化的短期过程进行预测。更多还原

【Abstract】 Internet is gradually becoming the new forces of social opinion propagation. Public opinion on the Internet can in large measure reflect the real condition of macroscopic consensus, so it has caused wide public concern of governments and enterprises. Recent years, it is becoming the hotspot in the domain of both social sciences and natural sciences that how public opinion on the Internet has impacted or will impact on real society and how to predict or guild the opinion effectively. However, most presented models do not consider the influence of network topology although they can explain some phenomenon occurred in the evolution of opinion well. These models can not be used to predict opinion evolution tendency.In this thesis, network opinion related issues are initially studied in pace with the demand for opinion evolution behaviors, prediction technologies and guiding mechanisms, and main contents include the influences of network topology characteristic on topic propagation and viewpoint exchange, the feature of network opinion evolution, and the prediction of opinion evolution tendency. The research work of this thesis is supported by Cultivating Fund Project for Major State Projects, University Technological Innovations Plan - ’Research on key technologies of internet opinion propagation and warning’ (No. 707006). The main innovations in this thesis are outlined as following:1. The influence of network topology on propagation of topic is studied, and the effects of some kinds of networks, which include completely connected network, small world network and scale-free network, on propagation velocity of topic and stability of community consensus are mainly inspected. It is also considered that the relationship between the number and distribution characteristic of individuals holding consensus in an initial stage and the propagation process. Some helpful conclusions for opinion guidance are reached. For example, heterostructure of nodes is a deterrent to topic propagation when average degree of nodes is larger; increasing clustering coefficient of network can pick up propagation speed; there is mutual check between network topology characteristics which influence topic propagation speed.2. The influence of network topology on process of consensus exchange is researched. The definition of parameter, which represents the dispersion degree of individual consensus, and parameter used to describe the changing tendency of community opinion, are introduced. The principal conclusions include, small world property and homojunction of network can lead to enough information exchange which is benefit for reaching unified consensus; scale-free property of network can prolong the process of consensus polymerization which leads to the condition that minority will exist throughout; for networks with heterostructure higher clustering level can speed up viewpoint exchange process and shorten the time that opinion reaches at steady state.3. Some features of process of network opinion evolution are analyzed. It is found that variant stages of topic development show homogeneity in a sense. If consider this homogeneity as the inertia of evolution system, the existence of inertia should be an important basis for the prediction of evolution tendency. Further more, the hits and deployment time of hot topic have correlation, which has a bearing on the tendency of the whole series and the selection to opinion guiding methods.4. Prediction methods to the tendency of network opinion evolution are also initially studied and the way to predict network opinion tendency based on time series analysis theories is present. The results of data analysis and empirical work show that the opinion evolution tendency for short periods can be able to predict by using the method of time series analysis.更多还原