社会网络中重要节点的挖掘研究

发布时间：2019-01-21 10:19

【摘要】：社会网络从根本上说是一种复杂网络,是复杂网络在现实世界中的应用。社会网络是个体与不同社会关系的集合,反映了社会个体之间的相互活动。社会影响力分析是社会网络的重要研究方向,其中个体的影响力和重要性的度量是一个热点问题,在抑制疾病传播和商品营销相关领域有着重要作用。目前已经有多种算法被提出,用来分析节点自身属性和特点,衡量个体在网络中的重要程度,本文主要研究社会网络中重要节点挖掘问题。考虑到已有的节点影响力和重要性度量算法大多只关注了节点的自身属性,显然节点的重要性不仅与其自身的局部属性相关,也与节点所在网络中的全局特性相关。基于此本文提出了一种基于双尺度的重要节点挖掘算法(KShell and Local Degree Centrality,KSLD),结合了节点的局部信息和全局信息。本文首先利用KShell分解对网络层次进行了划分并使用熵衡量节点影响全局网络层的能力,接着在局部信息的获取上由于传统中心性算法各有优劣,本文对传统算法的效率和效果做出平衡和改进,提出局部度中心性的算法计算节点局部信息,最后的实验对于局部度算法的效果与KSLD的应用分别做了论证和对比,并证明了结合了双尺度的KSLD算法对不同类型的网络的适用性比传统算法要好。目前大多数节点影响力的分析都是以静态网络为基础进行,而现实中网络是不断演变的,所以静态网络基础上的节点影响力特征并不能很好的适用于实际网络状况,针对这一问题,本文将个体影响力分析深入到演化网络中来,研究了演化网络中的信息传播特征,分析节点的影响力扩散方式,发现节点在网络演化中不断离开社区和加入新社区的事件属性,本文对于事件进行了详细的定义,并结合这一特点给出了一种基于事件的节点影响力评价方法(Based on Event Centrality,BE),通过社交指数和影响力指数两个指标对节点影响力进行了评估和排序找出影响力较大的重要节点,最后通过实验验证了该方法的可行性。
[Abstract]:Social network is a kind of complex network, which is the application of complex network in the real world. Social network is a collection of individual and different social relations, reflecting the interaction between social individuals. Social impact analysis is an important research direction of social network, in which the measurement of individual influence and importance is a hot issue, and plays an important role in the field of disease suppression and commodity marketing. At present, a variety of algorithms have been proposed to analyze the attributes and characteristics of nodes and to measure the importance of individuals in the network. This paper mainly studies the mining of important nodes in social networks. Considering that most of the existing algorithms only focus on the properties of nodes, it is obvious that the importance of nodes is not only related to their own local attributes, but also to the global characteristics of the network in which the nodes are located. Based on this, an important node mining algorithm, (KShell and Local Degree Centrality,KSLD, is proposed, which combines the local and global information of nodes. This paper first uses KShell decomposition to divide the network layer and uses entropy to measure the ability of nodes to influence the global network layer. Then the traditional centrality algorithm has its own advantages and disadvantages in obtaining local information. In this paper, the efficiency and effect of the traditional algorithm are balanced and improved, and the local degree centrality algorithm is proposed to calculate the local information of nodes. Finally, the effect of the local degree algorithm is proved and compared with the application of KSLD. It is proved that the two-scale KSLD algorithm is more suitable for different types of networks than the traditional algorithm. At present, most of the analysis of node influence is based on static network, but in reality the network is constantly evolving, so the characteristics of node influence based on static network can not be well applied to the actual network situation. In order to solve this problem, this paper studies the characteristics of information dissemination in evolutionary networks, and analyzes the influence diffusion mode of nodes. In this paper, the event attribute of node leaving community and joining new community is found in the evolution of network. In this paper, the event is defined in detail, and an event based node impact evaluation method, (Based on Event Centrality, is given. BE), evaluates the influence of nodes by social index and influence index, and sorts out the important nodes with great influence. Finally, the feasibility of this method is verified by experiments.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O157.5

【参考文献】