社交网络中个体交互行为和观点演化模式研究

发布时间：2018-03-19 15:05

  本文选题：社交网络　切入点：网络舆论　出处：《北京交通大学》2016年博士论文　论文类型：学位论文

【摘要】：随着互联网在全球范围内的飞速发展,网络媒体被公认为是继报纸、广播、电视之后的“第四媒体”,网络已成为反映社会舆情的主要载体之一。社交网络是网民进行信息交流的重要平台,其独特的组织模式催生出网络生态的复杂化和观点传播的日趋碎片化,是网络舆论形成和传播中不可忽视的力量。社交网络具有便捷多样的接入方法和更为灵活的交互参与方式,同时包含用户个体的自主性、交互行为的复杂性和群体的异质性等潜在特征。网络舆论的形成、演化和传播依赖于用户之间的交互行为,然而传统舆论研究难以真实反映社交网络舆论中的用户特性,由此,一系列针对网络舆论中个体交互行为和观点演化模式研究的新挑战日渐凸显：个体交互研究需要多角度、细粒度地刻画其行为特性,例如用户关系网络结构、交互决策的选择偏好：观点演化研究应重视与用户影响力和影响范围的结合；舆论传播演化研究应体现自由观点信息之间的分化竞争等。鉴于此,本文结合交叉学科研究方法,围绕社交网络中个体交互行为,对社交网络用户数据采集与用户特性实证分析、用户交互行为建模、观点交互机制与演化过程、竞争信息传播模式等问题进行了深入研究。论文采用实证数据分析、数学建模和计算机仿真相结合的研究方法,重点研究社交网络中个体交互和观点演化的微观规则和宏观规律,探索网络舆论的形成机制,刻画网络舆情的演化过程。研究不仅能够加深对复杂系统和网络舆情本质的认识,丰富社交网络中个体交互行为、网络观点交互与演化和舆论形成与传播等方面的理论,而且能够在网络舆论的及时应对、合理引导等方面得到实际应用。论文的研究工作得到了国家自然科学基金项目(No.61271308、No.61401015)、北京市重点实验室资助项目和北京市重点学科建设项目等项目的支持。论文的主要工作和创新点如下：1.单向关注关系是微博不同于其他社交网络的关系结构特征,改变了用户获取信息的方法和传播信息的途径,论文实证分析微博数据,进而发现单向关注对微博关系网络结构、用户交互行为和个体偏好带来影响。研究中采用基于网络爬虫的数据采集框架获取大量微博用户数据,实证分析发现：单向关注是导致节点度分布幂律偏移的主要原因,双向关注在用户好友关系中所占比例较低,呈现负指数衰减特征;用户交互行为存在明显偏好,单向关注和双向关注用户转发行为表现出随用户偏好幂律衰减特性；双向关注关系尽管所占比例低,却表现出高频度的交互行为和优先的用户偏好,在用户交互过程中影响显著。研究加深对社交网络中用户关系和行为特性的认识,并为网络舆论中的微观个体研究提供理论依据和数据支撑。2.用户交互行为在网络舆论的形成过程中起重要作用,多方面研究已经发现了个体复杂行为中的幂律特性和多模特性,然而却难以采用数学方法系统地描述并解释相关现象,论文建模微观个体交互行为,描述并研究社交网络用户复杂交互行为及其时间特性。论文在实证分析用户选择偏好特性的基础上,提出偏好优先的兴趣信息决策机制,建立基于选择偏好的社交网络用户交互模型,研究分析观察范围和兴趣概率在微观个体交互作用中的具体表现,最后通过微博用户数据验证模型的有效性。研究发现交互环境的改变将直接导致不同模式的交互结果,固定观察范围对应于交互行为的负指数特性,随机观察范围则对应于幂律特性。论文分析了兴趣信息集合中偏好用户的随机化表现,解释了交互行为分布中平稳期、衰减期和截断期的成因,并给出了个体交互行为及其多模特征的数学描述方法,量化了在面对大量信息流时选择偏好对交互行为的直接影响。论文研究能够真实反映社交网络中个体交互现象,深化了对微观交互规律的认识,具有重要的理论和实际意义。3.传统观点交互研究局限于相邻个体之间的简单交互规则,难以准确描述社交网络中舆论形成与演化过程,论文采用观点动力学建模方法,研究广阔的信息获取范围和个体异质性对舆论演化的影响。论文提出基于扩展观察范围与社会影响力的观点交互和演化模型,借鉴CODA思想描述用户个体观点；结合多数群体和少数群体的社会作用,提出同时考虑用户观点强度、观点接触直接性和作用用户数量的群体观点交互机制,扩展了用户个体的观察范围：并引入成功劝说作用的影响力反馈累积机制,使得话题讨论与观点演化更加符合社交网络中的实际特征。研究发现了个体交互强度和观点稳定性之间存在制衡关系,以及影响力逐渐累积表现出的异质性特征。该方法建立起微观个体交互和宏观舆论演化的纽带,结合用户影响力和影响范围,量化了突破局部邻居的观点交互作用,能够为社交网络中追踪舆论动态、分析网络舆情等应用领域提供理论支撑。4.信息之间的竞争和碰撞是社交网络中用户自由表达观点并广泛参与信息传播的结果,而目前国内外对多元竞争信息之间的交互作用和传播模式鲜有研究,论文建立竞争信息传播模型,研究社交网络中多元竞争信息的交互作用和传播演化过程。论文研究借鉴传染病动力学理论,抽象出网络用户面对竞争信息时的四种状态,提出基于二元竞争信息交互的SHIR传播模型。研究中利用平均场理论,推导出刻画节点传播状态随时间演化的微分方程组。仿真分析表明：中立的犹豫状态催化了话题在传播过程中的充分讨论,提高了整体信息的总覆盖率；二元信息的竞争作用表现为对内部有限用户资源的争抢；优势信息主导了信息的传播与演化过程,竞争信息的终态信息密度比值与其稳态转移概率比值存在幂律关系。研究建立了竞争信息交互与多元信息传播之间的联系,弥补了现有研究中信息交互与传播分离的不足之处,为准确研究分析多维度的复杂网络舆情提供了切实可行的研究方法。
[Abstract]:With the rapid development of the Internet in the global scope, the network media is recognized as the newspapers, radio, television after the "fourth media", the network has become the main carrier of public opinion. One of the social network is an important platform for Internet users to exchange information, its unique organizational model spawned increasingly complex and fragmented the views of network ecology, is a force to be reckoned with the network of public opinion formation and dissemination. Social network has a variety of convenient access method is more flexible and interactive participation, also include autonomy of the individual users, such as the complexity and heterogeneity of the potential characteristics of group interaction behavior. The formation of the network of public opinion the evolution and propagation depends on the interactive behavior between users, but the traditional public opinion research cannot accurately reflect the characteristics of users, social network public opinion in the network and a series of needle The new challenge mode evolution of individual interactions views and actions become increasingly prominent: the individual interaction research needs multi angle, fine-grained description of its behavior, such as the user relationship network structure, the preference of interactive decision making: a study of the viewpoint of evolution should be combined with the user attention and influence the scope and impact of the evolution of public opinion should be reflected; between the view of freedom of information differentiation competition. In view of this, this paper combined with the interdisciplinary research methods, focuses on individual interactions in social networks, analysis of social network user data acquisition and user characteristics of empirical, user interaction modeling, view interactive mechanism and evolution process, in-depth study of competition mode of information communication and other issues. The the empirical analysis, research methods of mathematical modeling and computer simulation combined with the focus on social network interaction and individual view The micro and macro point rule evolution rule, explore the formation mechanism of the network public opinion, modeling the evolution of network public opinion. The research can not only deepen our understanding of the complex system and the essence of network public opinion, rich individual social network interaction behavior, network interaction and evolution of ideas and public opinion formation and dissemination of the theory, but also in a timely manner with the network of public opinion, the practical application of a reasonable guide and so on. The research work of this thesis is supported by the National Natural Science Fund Project (No.61271308, No.61401015), supported by the Key Laboratory of Beijing city and Beijing city of key discipline construction projects such as support. The main work and innovation are as follows: 1. the one-way attention relationship structure micro-blog features different from other social networks, has changed the way of user access to information and dissemination of information, the empirical analysis Micro-blog data, and then found unidirectional interest in micro-blog network structure, the impact of user interaction and individual preference. The acquisition of a large number of users of micro-blog data collection framework web crawler based on the empirical analysis found that the one-way attention is the main cause of the node degree distribution of power-law offset, two-way attention for users of friends in a low proportion, negative exponential decay characteristics; there are obvious preference of user interaction, one-way and two-way attention focused on user behavior with forwarding user preference power-law attenuation characteristics; double to concern the relationship despite low proportion, but showed a high frequency of interactions and the priority of the user preferences, significant effects in the user in the interactive process. The research to deepen understanding of the social network user relationship and behavior characteristics, and provide rationale for the study of the individuals in the Internet public opinion On the basis of theory and data support.2. user interactions play an important role in the formation process of the network of public opinion, many studies have found that the power-law characteristics and characteristics of individual behavior in the multimode complex, but it is difficult to use mathematical methods to describe and explain the related phenomenon, the modeling of individual interaction behavior, describe and study social complex network user interaction behaviors and time characteristics. In the empirical analysis based on the selection of user preference characteristics, proposed interest preference information in decision-making mechanism, establish social network user interaction model selection based on preference, research and analysis of specific performance observation range and interest probability in the micro individual interaction, finally through the effectiveness of micro-blog users to validate the model. The study found that interactive environment changes will lead directly to the interaction results of different models, fixed observation range The negative exponential characteristics corresponding to the interaction behavior of the random observation range corresponds to the power law. This paper analyzes the performance of user preference information in randomization in the collection, explains the interactive behavior of the stable distribution, attenuation and truncation cause period, and describe the method of individual interaction behaviors and characteristics of multimode mathematical quantification in the face of a lot of information flow preferences directly influence on the interactive behavior. This study can reflect the phenomenon of individual interactions in social networks, to deepen the understanding of the microscopic interaction rules, simple interaction rules has important theoretical and practical significance of.3. traditional view research was limited to the adjacent interaction between individuals, it is difficult to accurately describe the process with the evolution of public opinion formation in social networks, the opinion dynamics modeling method of wide scope of information acquisition and individual heterogeneity of public opinion The influence of evolution. In this paper extended view interaction and evolution model of the observation range with the influence of society based on the thought of using CODA to describe the individual user view; combined with the social role the majority and minority groups, put forward considering the user views the strength, viewpoint of population interaction mechanism and the role of direct contact point of the number of users, the expansion of the individual users the observation range and the introduction of successful persuasion: the influence of cumulative feedback mechanism, makes the topic of discussion and opinion evolution accords with the actual characteristics of social network. The study found there was a relationship between the individual interaction strength and point stability, heterogeneity and cumulative influence gradually shows. The method to establish ties and individual interaction macro public opinion evolution, combined with the user influence and scope, to quantify the local neighborhood point to break through The interaction of public opinion to track dynamic social network analysis, competition and collision between the.4. information theory is a social network user freedom of expression and broad participation in the information dissemination network public opinion and other applications, while at home and abroad to study the interaction and communication between multiple few competitive information, the establishment of the competition information dissemination model, interaction and dissemination of information on diverse competition in social network evolution. Infectious disease dynamics theory research, abstract network users face four kinds of competition information, put forward the SHIR propagation model two yuan competition based on information interaction. Based on mean field theory, differential equations describes propagation state with time evolution is derived. The simulation analysis shows that the neutral state of hesitation in the dissemination process of the topic of catalysis The full discussion, to improve the overall information of the total coverage of two yuan; competition information for internal users for information resources; advantages dominated the propagation and evolution process of information, competition information terminal state information and the steady transition probability density ratio the ratio of the power law relation is established between the competitive information research. Interact with multiple communication links, make up for the deficiencies of the separation of information exchange and spreading in the present research, the research provides a feasible method for accurate analysis of the complex network of public opinion in various dimensions.

【学位授予单位】：北京交通大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP393.09
，

本文编号：1634791