社交网络中的信息与影响力传播模式研究

发布时间：2018-01-05 08:21

本文关键词：社交网络中的信息与影响力传播模式研究　出处：《北京交通大学》2017年博士论文　论文类型：学位论文

【摘要】：在当今互联网络时代大背景下,电视、广播、报纸等曾经主流媒体的地位正逐渐降低,而具有便捷性、实时性、低门槛等特点的网络媒体开始活跃起来。社交网络作为网络媒体的载体,在融合了社交娱乐、新闻传播、信息推广等多种元素的情况下,成为了人们接通外界过程中不可或缺的一扇窗户,也正是由于社交网络综合性强、结构错综复杂等特点,使得社交网路中的舆论相比于传统的舆论更加的复杂。不同社会群体间信息传递更加频繁,个体间的交互模式、参与话题及信息的方式更加多样化,而传统的舆论研究较难适用于上述的新环境。因此,探索社交网络信息传播规律、个体交互模式、特征差异性分析等问题的必要性日益凸显:研究需要从不同的角度分析用户行为,更为细致的刻画信息交互规律,比如不同情境下的用户决策行为模式和交互模型,同时,需要结合影响力与影响范围的分析,从社交网络的复杂交互情境入手,研究信息竞争与信息价值的内在关系。鉴于此,本文从多学科交叉的角度,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,围绕社交网络中信息与影响力的传播规律,对社交网络中的信息传播过程、信息干扰与竞争传播模式、用户影响力分析和影响力传播最大化等问题进行了深入的探索。本研究不仅能够帮助相关研究者加深对复杂网络研究和影响力问题的认识,丰富社交网络中个体交互行为演化和信息传播等方面的理论,而且能够为解决实际问题提供有效的帮助。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.内容可靠性是信息的重要特征属性,然而,在大多数社交网络信息传播研究中,缺乏对可靠性因素的系统建模与分析。研究结合了社交网络用户行为的主观性特点,分析了内容可靠性因素的作用机理,并根据社交网络信息的实际交互情景,提出了用户的反馈劝说机制,建立了基于内容可靠性因素的信息传播模型。通过建立平均场速率模型,并结合蒙特卡洛仿真实验,分析了用户对信息可靠性的怀疑程度与信息实际传播结果之间关系,并分析了在不同网络结构中可靠性影响的差异。仿真结果发现:信息内容可靠性因素对信息传播的实际结果起到决定性作用,用户对信息可靠性的怀疑程度显著影响着信息传播范围、传播速度、传播阈值、传播周期,而且该影响对网络疏密程度的依赖性较低;此外,发现信息传播的实际影响力为用户对信息内容可靠性的认可度,实际影响力的传播范围远小于信息的扩散范围。研究社交网络信息传播研究中加入信息内容可靠性因素,能够加深对信息传播动力的认识,丰富复杂网络理论,为探索社交网络信息传播及演化规律提供了有效的帮助。2.衍生信息传播干扰在社交网络信息传播过程中普遍存在,而传统的信息干扰模型仅从宏观角度分析信息传播及作用关系,缺乏对社交网络用户交互行为的微观刻画。根据社交网络用户与信息的交互特点,分析了衍生信息的产生条件、共存传播模式、用户与二元信息交互规则,并结合多元信息模型与社交强化理论,建立了一种基于衍生信息干扰的二元传播模型,针对衍生干扰现象,提出了先验干扰态与后验干扰态的概念,构建了二元社交强化规则,成功的将单信息模型中用户一维状态转化关系的扩展为二维状态转化关系。研究分析了基于衍生信息干扰的二元信息传播规律,并在规则网络与随机网络中进行了对比分析。实验结果表明:衍生干扰情况下的用户转发行为存在明显的"先入为主"现象,且干扰形式主要为后验干扰;此外,发现规则网络对信息干扰的时效性门槛要求比随机网络要低,这就表明规则网络下的信息传播更容易受到衍生干扰。研究衍生信息干扰现象与建模,有助理解社交网络多元传播模式中的网络干预现象,为信息干扰传播的研究提供新的研究思路,具有较高理论价值和实际意义。3.个体影响力分析是信息传播理论中个体异质性研究的最主要部分,效率与精度的权衡一直是社交网络个体影响力排序算法最主要问题。研究在社交网络拓扑特征的基础上,利用网络结构中影响力的传递特性,以节点中心性和权威性描述节点的影响力,并以中枢节点的连通性为核心,提出了关于社交网络节点影响力的迭代加权指标IEW(Iterative Equalization Weight)。研究分别采用皮尔逊系数相关性和网络鲁棒性分析方法,对比了 IEW指标与其他三类经典算法的优劣,并在真实网络数据集中采用信息传播动力学模型进行实际验证。验证结果表明:相比于传统算法,IEW通过牺牲一定运算速度获得了更高的准确性,并且算法的可靠性较有明显提高。研究提出的新算法为挖掘高效、可靠、准确的社交网络节点影响力评估算法提供了新思路,对深入研究社交网络信息与影响力传播规律起到理论支持作用。4.多节点的影响力传播最大化问题是结合信息传播理论与节点影响力分析的实际问题,传统算法难以根据实际需求灵活的调节算法的复杂度,并且算法可扩展性普遍较低。研究根据社交网络节点度幂律分布特性,分析并结合了贪心算法与启发式算法优缺点,提出了基于最优邻居发现的社交网络节点影响力最大化算法MNH(Max Neighbor Heuristic),该算法通过随机启发来构建贪心候选节点集,再利用计算节点边际增益的方式,近似估计最大影响力节点集合,实现了效率与精度的互换与调节。研究利用数学推导和理论证明的方式验证了算法的可行性和精确性,并在真实网络数据集中进行了蒙特卡洛仿真实验,利用独立级联传播模型与线性阈值传播模型,验证和对比了 MNH算法与其他三类经典算法的优劣。分析结果表明:虽然MNH算法的求解结果存在较为明显的波动性,但在算法平均耗时与精确度的综合分析上具有明显优势,并且表现出更高的可适性。研究较好的结合了启发式算法和贪心式算法的优点,为解决社交网络多节点传播影响力最大化问题的提供新方法与思路,是利用信息传播理论解决实际问题的一次有效尝试,具有较好的实际意义。
[Abstract]:In today's Internet era background, television, radio, newspapers and other mainstream media once status is gradually reduced, which is convenient, real-time, low threshold and other characteristics of the network media began to perk up. Social network as the carrier of the network media, the dissemination of news in the integration of social entertainment, information, promotion etc. elements of the case, to become the people on the outside of an integral part of the window, it is also due to the social network comprehensive structure, perplexing characteristics, makes social network public opinion compared to the traditional public opinion is more complex. Information transmission between different social groups more frequently, interaction between individuals, participation the topic and the way information is more diversified, and the traditional public opinion research is hard to apply to the new environment. Therefore, exploring the propagation rules of social network information, individual special interaction model. The necessity of syndrome differential analysis has become increasingly prominent problems such as: Study of user behavior analysis from different angles, more detailed characterization of the information exchange rules, such as the different situations user decision behavior model and interactive model, at the same time, it should be combined with the analysis of influence and impact range, starting from the complex situation of interactive social network study on the relationship between information, internal competition and the value of information. In view of this, this article from the multidisciplinary perspective, using empirical research methods of data analysis, mathematical modeling and computer simulation of combining the propagation and influence of information on social networks, the information dissemination process in social network, information interference and competition mode of transmission and probe into the influence of user analysis and the influence of the spread of the maximum problem. This study can not only help researchers deepen our understanding of the complex network Understanding of the problem and influence, rich individual behavior in social network evolution and dissemination of information theory, and can provide effective help for solving practical problems. 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 key project project support. The main work and innovation are as follows: 1. the content of reliability is an important attribute of information, however, in most of the social network information dissemination, lack of system modeling and analysis of the reliability factors. Based on the subjective characteristics of social network user behavior, analyzes the mechanism of content reliability factors and, according to the information of the actual social network interaction scenarios, proposes the user feedback mechanism is built based on persuasion, content Rely on the information dissemination model of factors. Through the establishment of the mean field rate model and Monte Carlo simulation, to analyze the relationship between the actual dissemination of results and information users suspected degree of reliability of information, and analyzes on the different network structure influences the reliability of difference. The simulation results showed that play a decisive role in the actual results of information content the reliability factors of information dissemination, users suspected degree of reliability of information significantly affects the scope of information dissemination, transmission speed, transmission threshold, propagation cycle, and the influence of dependence on network spacing is low; in addition, found that the actual influence of information dissemination for the recognition of the user of the information content of reliability, diffusion range the actual influence spread far less than information. With information content reliability factors research on social network information dissemination study, can add Deep understanding of dynamic information dissemination, enrich the theory of complex network, in order to explore the social network information dissemination and evolution provides effective help.2. derived information dissemination interference exists in the social network information dissemination process, information and the traditional interference model only from a macro point of view of information dissemination and interaction, the lack of social micro characterization network user interaction behavior. According to the interactive features of social network users and the information, analysis of the condition of producing derivative information, coexistence mode of transmission, and two yuan of user information interaction rules, and combining with multiple information model and social reinforcement theory, set up a two yuan propagation model derived information interference based on derivative interference phenomenon put forward the concept of state interference, prior and posterior state interference, constructing two yuan social enforces the rules, the success of the single user information model in one-dimensional. Extended state conversion between two-dimensional transformation relationship. Research and analysis of the derivative information interference two yuan of information dissemination based on rules and analyzed in regular networks and random networks. The experimental results show that the derivative interference case user forwarding behavior, there is obvious "phenomenon, and the interference First impressions are strongest" is mainly in the form of post in addition, checking interference; network rules for timeliness requirements are lower than the threshold of information interference random network, which shows the information dissemination rules under the network more vulnerable to interference. The derivative information derived interference and modeling, to help understand the network communication model of social network in the multi intervention phenomenon, to provide new research ideas for the study of information interference communication, it is of great theoretical value and practical significance of.3. analysis is one of the most influential individuals of individual heterogeneity in the theory of information dissemination The trade-off between efficiency and precision, has been the main problem of social network influence individual sorting algorithm. Based on social network topological features, the transfer characteristics of influence in the network structure, the center node and authoritative description of node influence, and connectivity of the central node as the core, we propose an iterative weighted the IEW index on the social network node influence (Iterative Equalization Weight). Was analyzed by Pearson correlation analysis method and the network robustness, comparing IEW index and other three kinds of classic algorithms, and in the real network data set by the information transmission dynamics model is tested. Test results show that compared to the traditional algorithm. IEW obtained a higher accuracy by sacrificing certain speed, and the reliability of the algorithm is improved. The research is put forward A new algorithm for mining is efficient, reliable, and provides a new way for node influence social network evaluation algorithm is accurate, in-depth study of social network information and influence propagation to theoretically support a role for.4. multi node communication influence maximization problem is combined with the actual analysis of information transmission theory and node influence, the traditional algorithm is based on the actual needs the flexibility to adjust the complexity of the algorithm, and the algorithm scalability is generally low. According to the law of distribution characteristics of social network node degree power, combined with the analysis of the advantages and disadvantages of the greedy algorithm and heuristic algorithm, proposed the MNH social network node influence maximization algorithm based on optimal neighbor discovery (Max Neighbor Heuristic), the algorithm to construct a set of candidate nodes by random greedy heuristic, the computing node marginal gain, approximate the maximum impact of Festival The set of points, and the efficiency and accuracy of the exchange regulation. Research using mathematical derivation and theoretical proof to verify the feasibility of the algorithm and accuracy, and in the real network data set for the Monte Carlo simulation, using independent cascade propagation model and the linear threshold propagation model, validate and compare MNH algorithm with the other three classical algorithm. The results show that: Although the MNH algorithm for solving the volatility is more obvious, but the average time in comprehensive analysis and accuracy of the algorithm has obvious advantages, and showed higher applicability. On a good combination of heuristic algorithm and greedy algorithm, in order to provide with the idea of a new method to solve the maximization problem of the social network of multi node communication influence, is an effective attempt to solve practical problems by using the theory of information transmission, has better The practical significance.

【学位授予单位】：北京交通大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TP393.09;G206

【参考文献】