基于多层社会网络的信息扩散研究
发布时间：2018-09-12 07:50
【摘要】：社会网络是描述具有社会属性的个体及其相互间关系的一类网络,社会网络中信息扩散的研究已成热门领域,如舆情控制,可建立信息扩散模型,揭示其传播的特征及规律,再采用科学方法来预防和处理舆情的泛滥,由此可见,研究社会网络对真实社会网络有极为重要意义。随着网络信息爆炸式增长,社会网络中个体之间的关系不再是单一的,因此,近年来有研究者提出了一种新型的社会网络——多层社会网络(Multi-layered Social Network, MSN),即把复杂关系抽象成每层仅存一种社会关系的分层结构的社会网络,从而将平面网络关系图转换为立体网络结构,能够更好地刻画真实生活中社会网络形态与特征。 信息扩散过程中,首先将多层社会网络进行社团结构划分,其次建立多层社会网络的传播模型,最后采用影响最大化策略来实现信息扩散的最大化目标。多层社会网络是社会网络中最新的研究领域之一,社会网络中现存的很多算法和模型都不适用于多层社会网络,或者算法较粗糙等,因此,迫切需要解决这些关键问题。 本文针对多层社会网络信息扩散进行了如下研究： (1)多层社会网络的社团发现算法：社团是构成整个MSN的子集,其主要特征是每个社团内部的节点间的有着相对紧凑的连接方式,而各个社团之间却只存在着相对比较稀疏的连接方式。目前社团发现算法主要集中在单层的社会网络,而多层社会网络MSN的社团发现算法较少,社团划分结果较粗糙等特点,为了既考虑多层社会关系又区分对待不同层数的情况,以及考虑节点间本身的连接强度问题,本文提出了一种基于边聚类的多层社会网络社团发现(CLEDCC)算法。该算法充分考虑了层数给真实社会网络带来的现实意义,并对CLEDCC算法的数学模型和算法流程进行详细说明,仿真实验和分析,并与相关算法进行比较,该算法无需调整参数,算法稳定性高,划分结果精准。 (2)多层社会网络的信息传播模型：在单层社会网络中提出的信息传播模型,研究非常多,但是不太实用于多层社会网络,而目前这方面几乎没有相关研究。对于多层社会网络的处理一般都是通过模型简化处理,将多层复杂关系简化为单层社会网络,而单层社会网络的传播模型,对于问题的处理不够精细,故本文构建一种新型的多层电阻器传播(CRM)模型。详细重点地介绍该模型的理论基础,并由物理学电路模型的思想演变而来,通过图论、概率论等大量的数学理论做支撑,证明了模型的可行性。CRM传播模型有着非常好的特点,与真实网络的形态非常接近,CRM模型是非常适用于多层社会网络。 (3)信息扩散：给一个网络(V,E),在一定时间内,寻找尽可能少的种子节点,尽可能多的影响网络中的其他节点,这是一个NP-hard问题。采用本文提出的CLEDCC算法来实现更好的社团结构,其次利用CRM模型来仿真社会网络信息扩散过程,再扩展现有的社团与度启发CDH策略,通过CDH-CLEDCC来实现信息扩散最大化。
[Abstract]:Social network is a kind of network describing individuals with social attributes and their relationships. The study of information diffusion in social network has become a hot field. For example, public opinion control, information diffusion model can be established to reveal the characteristics and laws of its propagation, and then scientific methods can be used to prevent and deal with the overflow of public opinion. Thus, the study of society can be seen. With the explosive growth of network information, the relationship between individuals in social networks is no longer single. Therefore, in recent years, researchers have proposed a new type of social network, multi-layered social network (MSN), which abstracts complex relationships into each layer only. A social network with hierarchical structure of social relations is preserved, thus transforming the plane network diagram into a three-dimensional network structure, which can better describe the form and characteristics of social networks in real life.
In the process of information diffusion, the multi-layer social network is divided into several groups, and then the propagation model of multi-layer social network is established. Finally, the maximization of information diffusion is achieved by using the strategy of influence maximization. T-type is not suitable for multi-layer social networks, or the algorithm is rough, so it is urgent to solve these key problems.
In this paper, the information diffusion of multi-layer social networks is studied as follows:
(1) Club discovery algorithm in multi-layer social networks: Club is a subset of the whole MSN, and its main feature is that there is a relatively compact connection between nodes within each community, while there is only a relatively sparse connection between the various communities. Multi-layer social network MSN has fewer community discovery algorithms and rough community partitioning results. In order to consider the multi-layer social relations and different layers, and consider the connection strength between nodes, this paper proposes a multi-layer social network community discovery algorithm based on edge clustering (CLEDCC). Considering the practical significance of the real social network brought by the number of layers, the mathematical model and algorithm flow of CLEDCC algorithm are described in detail, and the simulation experiment and analysis are carried out. Compared with the related algorithms, the algorithm does not need to adjust parameters, and has high stability and accurate partition results.
(2) Information dissemination model of multi-layer social network: The information dissemination model proposed in single-layer social network has been studied very much, but it is not very practical for multi-layer social network, and there is little research in this area at present. A new multilayer resistor propagation (CRM) model is constructed in this paper. The theoretical basis of the model is introduced in detail, which is derived from the idea of physical circuit model and supported by a large number of mathematical theories such as graph theory and probability theory. The CRM communication model has very good characteristics and is very close to the real network. The CRM model is very suitable for multi-layer social networks.
(3) Information diffusion: For a network (V, E), it is a NP-hard problem to find as few seed nodes as possible within a certain period of time and to influence as many other nodes as possible in the network. The CLEDCC algorithm proposed in this paper is used to achieve a better community structure. Secondly, the CRM model is used to simulate the process of information diffusion in the social network, and then extended. The existing community and degree heuristic CDH strategy can maximize information diffusion through CDH-CLEDCC.
【学位授予单位】：陕西师范大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.0
本文编号：2238376
[Abstract]:Social network is a kind of network describing individuals with social attributes and their relationships. The study of information diffusion in social network has become a hot field. For example, public opinion control, information diffusion model can be established to reveal the characteristics and laws of its propagation, and then scientific methods can be used to prevent and deal with the overflow of public opinion. Thus, the study of society can be seen. With the explosive growth of network information, the relationship between individuals in social networks is no longer single. Therefore, in recent years, researchers have proposed a new type of social network, multi-layered social network (MSN), which abstracts complex relationships into each layer only. A social network with hierarchical structure of social relations is preserved, thus transforming the plane network diagram into a three-dimensional network structure, which can better describe the form and characteristics of social networks in real life.
In the process of information diffusion, the multi-layer social network is divided into several groups, and then the propagation model of multi-layer social network is established. Finally, the maximization of information diffusion is achieved by using the strategy of influence maximization. T-type is not suitable for multi-layer social networks, or the algorithm is rough, so it is urgent to solve these key problems.
In this paper, the information diffusion of multi-layer social networks is studied as follows:
(1) Club discovery algorithm in multi-layer social networks: Club is a subset of the whole MSN, and its main feature is that there is a relatively compact connection between nodes within each community, while there is only a relatively sparse connection between the various communities. Multi-layer social network MSN has fewer community discovery algorithms and rough community partitioning results. In order to consider the multi-layer social relations and different layers, and consider the connection strength between nodes, this paper proposes a multi-layer social network community discovery algorithm based on edge clustering (CLEDCC). Considering the practical significance of the real social network brought by the number of layers, the mathematical model and algorithm flow of CLEDCC algorithm are described in detail, and the simulation experiment and analysis are carried out. Compared with the related algorithms, the algorithm does not need to adjust parameters, and has high stability and accurate partition results.
(2) Information dissemination model of multi-layer social network: The information dissemination model proposed in single-layer social network has been studied very much, but it is not very practical for multi-layer social network, and there is little research in this area at present. A new multilayer resistor propagation (CRM) model is constructed in this paper. The theoretical basis of the model is introduced in detail, which is derived from the idea of physical circuit model and supported by a large number of mathematical theories such as graph theory and probability theory. The CRM communication model has very good characteristics and is very close to the real network. The CRM model is very suitable for multi-layer social networks.
(3) Information diffusion: For a network (V, E), it is a NP-hard problem to find as few seed nodes as possible within a certain period of time and to influence as many other nodes as possible in the network. The CLEDCC algorithm proposed in this paper is used to achieve a better community structure. Secondly, the CRM model is used to simulate the process of information diffusion in the social network, and then extended. The existing community and degree heuristic CDH strategy can maximize information diffusion through CDH-CLEDCC.
【学位授予单位】：陕西师范大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.0
【共引文献】
相关期刊论文 前2条
1 赵炎;郑向杰;;网络嵌入性与地域根植性对联盟企业创新绩效的影响——对中国高科技上市公司的实证分析[J];科研管理;2013年11期
2 陈柏彤;张斌;;科学知识扩散研究框架[J];图书情报工作;2014年15期
相关博士学位论文 前3条
1 吴联仁;基于人类动力学的社交网络信息传播实证分析与建模研究[D];北京邮电大学;2013年
2 张莉莉;组织文化对于组织成员作用机制研究：借鉴催化动力学方法[D];北京交通大学;2012年
3 郑向杰;基于企业嵌入视角的联盟创新网络中知识共享与创新研究[D];上海大学;2014年
，本文编号：2238376
本文链接：https://www.wllwen.com/guanlilunwen/ydhl/2238376.html
