基于社区结构的网络影响力传播算法研究

发布时间：2018-10-26 08:01

【摘要】：近年来,随着互联网技术的日益普及,人们获取信息的途径也在发生着悄然变化,从最初的广播电视报纸,到如今的微博贴吧朋友圈。在线社会网络不断与传统人际网络相融合,产生海量数据,为社会网络分析带来了前所未有的机遇。大批科研工作者对社会网络影响力最大化、信息传播规律等课题进行深入研究与分析。其中,如何选择社会网络中最具影响力的前K个节点及如何构建准确的信息传播模型这两个方向,成为了社会网络领域的一个研究热点。本文首先在深入分析前人研究的基础上,针对现存社会网络影响力最大化算法所存在的问题,引入弱连带优势理论提出了一种改进的中心性算法。其次,详细分析了线性阈值模型,结合社会网络不同节点间的差异性,引入节点关联强度的概念及信息自身的引力特性,提出一种新型的社会网络传播模型。具体的研究内容如下:(1)基于社区结构的关键节点中心性算法。结合网络社区结构特性,将边界节点及社区内部节点同时作为关键节点,来衡量其实际影响力。根据弱连带优势理论,内聚性很强的网络并不利于节点获取外部信息,因此考察连接不同社区的弱连带关系即边界节点的属性,有利于信息的跨区域传播;同时,选取社区内部最具影响力节点可以使信息在社区内快速传播。二者结合有利于信息在全网中的扩散。本文从3个不同方面分别在3个数据集上验证了算法的有效性。(2)关联强度阈值模型。通过对线性阈值模型的深入研究发现,该模型中假设某一节点同一时刻受到来自其邻接节点的影响力值均相同。然而,在真实社会网络中,不同个体间存在着远近亲疏的关系,个体的差异性决定了节点受其邻居节点影响的差异性;同时,信息的传播效果与信息自身的吸引力密切相关。因此,本章提出一种基于LT模型的——关联强度阈值模型。该模型通过吸收线性阈值模型的优点,结合社会网络不同节点间的差异性,引入节点关联强度的概念及信息自身的引力特性,对线性阈值模型中的参数做了改进并提出了新的节点影响力的计算公式。
[Abstract]:In recent years, with the increasing popularity of Internet technology, people's access to information is also quietly changing, from the original radio and television newspapers, to today's Weibo post bar friends. The combination of online social network and traditional interpersonal network produces massive data and brings an unprecedented opportunity for social network analysis. A large number of researchers deeply study and analyze such topics as maximization of social network influence and rules of information dissemination. Among them, how to select the most influential first K nodes in social network and how to build an accurate information dissemination model have become a research hotspot in the field of social network. In this paper, based on the analysis of previous studies, an improved centrality algorithm is proposed by introducing the weak joint advantage theory in order to solve the problem of the existing algorithms for maximizing the influence of social networks. Secondly, the linear threshold model is analyzed in detail, and a new social network propagation model is proposed by introducing the concept of node association strength and the gravitational properties of information itself, combining with the difference between different nodes of social network. The specific research contents are as follows: (1) the key node centrality algorithm based on community structure. Combined with the characteristics of the network community structure, the boundary node and the community internal node are taken as the key nodes simultaneously to measure their actual influence. According to the theory of weak joint advantage, the strong cohesion of network is not conducive to the node to obtain external information, so the study of the weak link between different communities, that is, the attributes of the boundary node, is conducive to the cross-regional dissemination of information. At the same time, selecting the most influential nodes in the community can make the information spread quickly in the community. The combination of the two is beneficial to the diffusion of information in the whole network. This paper verifies the validity of the algorithm on three data sets from three different aspects. (2) the threshold model of association strength. Through the in-depth study of the linear threshold model, it is found that the model assumes that a node is affected by the same value from its adjacent nodes at the same time. However, in the real social network, there are close and distant relationships between different individuals, and the difference of individuals determines the difference of nodes affected by their neighbors. At the same time, the communication effect of information is closely related to the attraction of information itself. Therefore, in this chapter, an association strength threshold model based on LT model is proposed. By absorbing the advantages of the linear threshold model and combining the differences between different nodes in the social network, this model introduces the concept of node association strength and the gravitational properties of the information itself. The parameters in the linear threshold model are improved and a new formula for calculating nodal influence is proposed.
【学位授予单位】：哈尔滨工程大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP301.6

【参考文献】