在线社交网络模型演化及传播机制研究

发布时间：2018-01-15 23:14

本文关键词：在线社交网络模型演化及传播机制研究　出处：《山东师范大学》2016年博士论文　论文类型：学位论文

【摘要】：随着全球互联网技术的发展,在线社交网络已经取代了电视、报纸等传统媒体,成为主流信息传播平台。信息传播方式的改变,使人们在快捷获取、传播信息的同时,也受到了信息爆炸、虚假谣言等问题的干扰。因此,研究并发现信息在社交网络的传播规律,对互联网的信息管理、舆情管控、网络事件的实时跟踪及预警等具有重要作用,对维护国家安全及社会稳定也具有十分重要的现实意义。基于此,本文针对社交网络的演化机制及信息传播的相关理论展开研究,主要内容如下:(1)在BA无标度网络模型基础上,建立基于边数随机增长的网络演化模型,尝试揭示真实网络度分布在双对数坐标下头部弯曲现象的机理。网络演化过程中,节点在加入社交网络时选择朋友的数目具有随机性,本文根据边数增长数量完全服从泊松分布、部分服从泊松分布和选择概率完全随机三种情况,建立了三个网络演化模型,并推导出三个模型的度分布函数。研究表明,在保持择优概率的机制下,边数增长的随机性可以导致度分布出现弯曲现象;当节点度较大时,度分布仍然服从幂律分布;如果将择优概率变成随机概率,则会破坏度分布的幂律分布。根据模型算法,本文构建了相应的仿真网络模型,实验数据验证了模型假设的合理性和理论分析的正确性。(2)根据在线社交网络信息传播特点和目前社交网络传播模型研究中存在的问题,提出一种基于用户相对权重的信息传播模型——RWSIR模型。社交网络中,信息能否顺利传播,很大程度上取决于传播者和接收者双方的地位关系,本文定义了社交网络用户间的相互影响力函数,并对网络中的传播路径及传播过程进行了分析,讨论了不同路径的传播影响力,提出了一种基于相对权重的信息传播模型,给出了传播动力学方程。为进一步验证模型的有效性,本文在不同的网络拓扑结构下,分别以权威节点和普通节点作为传播节点,将传统的SIR模型和本文模型进行了仿真实验。实验数据表明,普通节点在传播过程中一般会弱于权威节点,而在某些特殊条件下差别不大;两类模型在均匀网络中没有明显差异,但在非均匀网络中存在明显差异,本文模型更能体现真实网络中信息传播的特点。(3)根据谣言在社交网络中的传播特点,基于SEIR传染病模型提出了一种改进的谣言传播模型谣言传播不同于人类的传染病传播,民众听到谣言后会首先辨别其真伪性,然后决定是否传播。本文根据民众对谣言的掌握情况将整个网络中的人群分为四类:未知消息者、传播者、知情者、不感兴趣者。为详细刻画民众接收谣言后的不同反应,引入了四类人群的不同转化概率,建立了谣言传播模型动力学方程,并证明了方程解的稳定性,分析了不同转化概率对方程解的影响。最后在不同的网络拓扑上进行了仿真实验,实验数据验证了理论的正确性。(4)针对静态网络节点影响力排序算法的局限性,提出一种基于传播概率的节点影响力算法——PIC算法现有研究表明,网络中节点的影响力随着传播概率的不同而发生变化,基本节点影响力排序算法不能很好的解决上述问题。据此,本文在分析传播节点对网络中不同层邻居节点的影响力基础上,考虑传播概率对信息传播的影响,构建了一种既考虑网络拓扑结构,又体现传播概率的节点影响力算法,并进一步给出了近似计算方法。本文利用基本算法和PIC算法计算了不同网络的节点排序,数据结果显示,本文提出的算法综合稳定性和效率要优于其余基本算法。
[Abstract]:With the development of Internet technology, online social network has replaced the television, newspapers and other traditional media, has become the mainstream of information dissemination platform. Changes in the way of information dissemination, so that people in the fast access, dissemination of information at the same time, is also affected by the explosion of information, false rumors and other issues. Therefore, to study and find out the propagation law of the information in the social network, information management, the Internet public opinion monitoring, network event tracking and warning plays an important role, but also has very important practical significance to safeguard national security and social stability. Based on the relevant theories of the evolution mechanism for social networks and information dissemination, main content are as follows: (1) the scale-free network model based on BA, the establishment of evolution model of randomly growing edges based on the network, try to reveal the true network degree distribution in double logarithmic coordinates The mechanism of head bending phenomenon. The network evolution process, node selection in the number of friends to join the social network with random, according to the number of edges of the growth in the number of completely obeys the Poisson distribution, the Poisson distribution and the probability of choosing some completely random three cases, established the evolution model of three networks, and deduces the three models the degree distribution function. The results show that the preferred mechanism to keep the probability, random number of edges growth can lead to the degree distribution of bending phenomenon; when the node degree is large, the degree distributions follow the power-law distribution; if the preferred probability into random probability, it will destroy the power-law degree distribution. According to the model algorithm in this paper, builds the corresponding network model, the experimental data to verify the correctness of the model assumptions and theoretical analysis. (2) according to the online social network information dissemination characteristics and the current society To research the problems in network communication model, put forward a kind of information communication model based on the user's relative weight RWSIR model. The social network, smooth information communication, depends largely on the status of both the disseminator and the receiver, this paper defines the mutual influence function of the social network between users, and to spread the path and propagation process in the network are analyzed, discussed the different path of the spread of influence, put forward a kind of information dissemination model based on relative weight, transmission dynamics equations are given. For further validation, based on different network topologies, with authority nodes and ordinary nodes respectively as communication node, the SIR model and the traditional model of the simulation experiments. The experimental data show that ordinary nodes in the communication process generally weaker than the authority of the node, and In some special conditions no difference; there is no significant difference in the homogeneous network two models, but there were significant differences in the heterogeneous network, this model can better reflect the characteristics of the information spreading in real networks. (3) according to the propagation characteristics of rumors in a social network, based on the SEIR epidemic model is proposed a rumor propagation model for rumor improved is different from the spread of infectious diseases of mankind, people after hearing the rumors will be the first to identify its authenticity, and then decide whether to spread. According to the people to grasp the rumors in the network group is divided into four categories: the unknown message, communicator, not feeling interested. For different reaction to describe people receive rumors after the different conversion probability introduced four kinds of people, to establish the dynamic equation of rumor propagation model, and prove the stability of the solution of equations, analysis of different Influence of probability of the integral equation. Finally, simulation experiments are carried out in different network topologies, the experimental data to verify the correctness of the theory. (4) aiming at the limitations of static network node influence sorting algorithm, proposed based on node transmission probability influence algorithm - PIC algorithm of existing research shows that the nodes in the network with the influence of propagation probability changes, the basic node influence ranking algorithm can't solve the problem. Therefore, based on the analysis of influence of different communication nodes based on the network layer of the neighbor nodes, considering the influence of the propagation probability of information dissemination, constructs a both network topology, node influence the algorithm shows the propagation probability, and further gives the approximate calculation method. This paper uses the basic algorithm and PIC algorithm in different network node scheduling, The results show that the integrated stability and efficiency of the proposed algorithm are better than those of the other basic algorithms.

【学位授予单位】：山东师范大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP393.09;G206

【相似文献】