基于复杂网络理论的网络核心边缘结构算法研究及应用

发布时间：2019-01-26 14:49

【摘要】：在线社交网络平台中,信息主体的行为活动多样性,一方面使得主体之间的相互关系变得复杂,另一方面也促进了信息的传播与扩散。基于复杂网络理论,本文旨在从海量社交网络数据中抽取信息主体之间的社交关系网络和行为关系网络,建立其复杂网络模型。在此基础上,从单层网络和多层网络视角,研究网络核心-边缘结构,并应用于识别核心节点在行为活动中作用。网络核心-边缘结构检测及其应用,将有助于提供更加优质的信息服务,并将产生更多的社会经济价值。本文的工作主要包含以下两个方面:1)基于新浪微博数据的复杂网络建模与分析。对爬取的微博数据进行提取和预处理,微博事件转发网络及其对应用户关系网络本身符合多层网络的特性。本文将从单层以及多层网络的视角出发,重点研究网络结构特征以及网络核心-边缘结构。研究在不同类型的事件转发过程中,参与转发的用户行为活动之间的关联性及其时序演化机制。通过研究多层网络中参与转发用户之间的行为关系来熟悉微博事件转发过程中的特点,包括分析转发网络度分布,网络随时间的演化趋势,网络结构的变化,事件转发网络与对应用户关系网络之间的关联性等。通过上述多方面的研究,更好地理解社交网络结构特征、信息主体之间的行为模式以及它们之间的相互作用关系。2)基于新浪微博数据的网络核心-边缘结构检测算法分析。在多关系社交网络系统中,首先尝试用不同的节点中心性方法对社交网络信息主体行为活动及其对应社交网络进行中心性分析,包括度中心性,接近中心性以及k-壳中心性。并结合社交网络基本特征,尝试用合适的方法对多层社交网络进行网络核心-边缘结构检测,理解不同方法的局限性。实验过程中发现,基于核层分解的核心-边缘结构检测算法能够有效地进行核心-边缘结构判定,它对核心节点的识别在大多数情况下也较为有效。
[Abstract]:In the online social network platform, the diversity of the behavior and activities of the information subject, on the one hand, makes the relationship between the agents become complex, on the other hand, it also promotes the dissemination and diffusion of information. Based on the theory of complex network, this paper aims to extract the social relationship network and behavioral relationship network between the information subjects from the massive social network data, and establish its complex network model. On the basis of this, the kernel-edge structure of the network is studied from the view of single-layer network and multi-layer network, and it is applied to identify the role of core nodes in behavior activities. Network core-edge structure detection and its application will help to provide higher quality information services and generate more social and economic value. This paper mainly includes the following two aspects: 1) complex network modeling and analysis based on Sina Weibo data. After extracting and preprocessing Weibo data from crawling, Weibo event forwarding network and its corresponding user relationship network accord with the characteristics of multi-layer network. In this paper, the characteristics of network structure and the core-edge structure of the network are studied from the point of view of single-layer and multi-layer networks. This paper studies the relevance and temporal evolution mechanism of user behavior activities involved in forwarding in different types of event forwarding processes. By studying the behavior relationship between the participating forwarding users in multilayer network, we are familiar with the characteristics of Weibo event forwarding process, including analyzing the distribution of forwarding network degree, the evolution trend of network with time, and the change of network structure. Correlation between event forwarding network and corresponding user relationship network. Through the above research, we can better understand the characteristics of social network structure, the behavior patterns among information agents and the interaction between them. 2) based on the data of Sina Weibo, the detection algorithm of core edge structure of network is analyzed. In the multi-relational social network system, we first try to use different node-centered methods to analyze the behavior of social network information subject and its corresponding social network, including degree centrality, proximity centrality and k-shell centrality. Combined with the basic characteristics of social networks, this paper tries to use appropriate methods to detect the core-edge structure of multilayer social networks and to understand the limitations of different methods. The experimental results show that the kernel-edge structure detection algorithm based on kernel layer decomposition can effectively determine the core-edge structure, and it is also more effective in the recognition of core nodes in most cases.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O157.5;TP393.09

【相似文献】