基于复杂网络分析的人物关系挖掘
发布时间：2018-11-14 15:26
【摘要】：真实世界的复杂系统通常可以抽象成节点和边构成的网络拓扑结构。随着对复杂系统的研究深入,复杂网络分析方法存在两方面问题。首先是复杂网络模型朝着异质化、多元化的方向发展。传统的复杂网络拓扑是复杂系统的高度抽象表达。随着研究深入,网络的关系异质性在网络研究问题越来越重要,如何对异质复杂网络进行算法分析是一个重要的研究方向。其次是网络规模越来越庞大。数据量的激增对复杂网络算法的存储和计算问题带来了严峻的挑战,能否从大规模网络拓扑提取一个近似的精简结构是一个重要的难题。为了解决上述问题带来的挑战,本文基于连边模式对复杂网络进行研究。从现有的研究成果显示,边模式有助于研究节点属性关系、网络生成模型、拓扑结构的高阶表达等的网络性质。本文利用边模式的研究方法并结合传统复杂网络分析理论,研究了复杂网络的关系异质性问题和核心结构表达问题。本文的主要贡献如下:1.本文提出了一种基于多层网络模型的重叠社团发现算法。本文系统地研究了连边社团检测(LCD)算法,这是一种单层网络下基于连边关系的重叠社团挖掘算法。本文基于原始算法的缺陷提出了改进算法,并且由于该算法在多层网络模型的适用性,提出了多层网络连边社团检测(MLCD)算法。该算法可用于异质关系的复杂网络模型。最后利用了社团性能检测的LFR框架,通过MLCD与主流的Louvain和Infomap社团发现算法结果进行实验对比,肯定了本算法的适用性和有效性。2.本文提出了一种复杂网络核心影响结构提取算法。该算法挖掘网络中每个节点邻域子图内的核心模体实例,然后将其合并构成核心影响结构。不同于传统核心结构挖掘方法,核心影响结构是一个精简的网络子图,它不仅包含了网络中的核心节点,还刻画了核心节点对非核心节点的影响关系。同时,该结构可以很好的体现原始网络的拓扑特征和尺度特征。该方法适用于网络参数估计、可视化分析等方面,同时也可以用于复杂系统的网络拓扑提取问题。综上所述,本文以边模式作为网络的基本对象,对复杂网络的关系异质性和核心影响问题进行了深入的研究,并且取得了有效的成果。所以,基于边模式的复杂网络分析方法可以作为未来复杂网络学科发展的重要研究工具。
[Abstract]:Real world complex systems can be abstracted into a network topology composed of nodes and edges. With the development of complex system, there are two problems in complex network analysis method. The first is the development of complex network model towards heterogeneity and diversification. Traditional complex network topology is a highly abstract representation of complex systems. With the deepening of the research, the relationship heterogeneity of network is becoming more and more important in network research, and how to analyze the algorithm of heterogeneous complex network is an important research direction. Second, the scale of the network is getting larger and larger. The rapid increase of data volume brings a severe challenge to the storage and computation of complex network algorithms. It is an important problem to extract an approximate reduced structure from large-scale network topology. In order to solve the challenge brought by the above problems, this paper studies the complex network based on the connected edge mode. It is shown from the existing research results that the edge pattern is helpful to the study of the network properties of node attributes, network generation models, and the higher-order representation of topological structures. In this paper, the relationship heterogeneity problem and the core structure representation problem of complex networks are studied by using the method of edge pattern and the traditional theory of complex network analysis. The main contributions of this paper are as follows: 1. In this paper, an overlapping community discovery algorithm based on multi-layer network model is proposed. In this paper, the (LCD) algorithm for community detection with connected edges is studied systematically, which is an overlapping community mining algorithm based on the link relation in a single-layer network. This paper proposes an improved algorithm based on the defects of the original algorithm, and because of the applicability of the algorithm in the multilayer network model, a new (MLCD) algorithm for community detection in multi-layer networks is proposed. The algorithm can be applied to complex network models of heterogeneous relationships. Finally, the LFR framework of community performance detection is used, and the results of MLCD are compared with those of the popular Louvain and Infomap community discovery algorithms, and the applicability and effectiveness of this algorithm are confirmed. 2. In this paper, an algorithm for extracting the core influence structure of complex networks is proposed. In this algorithm, the core motifs in the neighborhood subgraph of each node in the network are mined, and then combined to form the core influence structure. Different from the traditional core structure mining method, the core influence structure is a concise network subgraph, which not only includes the core nodes in the network, but also describes the relationship between the core nodes and the non-core nodes. At the same time, the structure can well reflect the topology and scale characteristics of the original network. This method is suitable for network parameter estimation, visual analysis and so on. It can also be used to extract the network topology of complex systems. To sum up, this paper takes the edge pattern as the basic object of the network, studies the relationship heterogeneity and the core influence of the complex network deeply, and obtains the effective results. Therefore, the analysis method of complex network based on edge pattern can be used as an important research tool for the development of complex network in the future.
【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
本文编号：2331562
[Abstract]:Real world complex systems can be abstracted into a network topology composed of nodes and edges. With the development of complex system, there are two problems in complex network analysis method. The first is the development of complex network model towards heterogeneity and diversification. Traditional complex network topology is a highly abstract representation of complex systems. With the deepening of the research, the relationship heterogeneity of network is becoming more and more important in network research, and how to analyze the algorithm of heterogeneous complex network is an important research direction. Second, the scale of the network is getting larger and larger. The rapid increase of data volume brings a severe challenge to the storage and computation of complex network algorithms. It is an important problem to extract an approximate reduced structure from large-scale network topology. In order to solve the challenge brought by the above problems, this paper studies the complex network based on the connected edge mode. It is shown from the existing research results that the edge pattern is helpful to the study of the network properties of node attributes, network generation models, and the higher-order representation of topological structures. In this paper, the relationship heterogeneity problem and the core structure representation problem of complex networks are studied by using the method of edge pattern and the traditional theory of complex network analysis. The main contributions of this paper are as follows: 1. In this paper, an overlapping community discovery algorithm based on multi-layer network model is proposed. In this paper, the (LCD) algorithm for community detection with connected edges is studied systematically, which is an overlapping community mining algorithm based on the link relation in a single-layer network. This paper proposes an improved algorithm based on the defects of the original algorithm, and because of the applicability of the algorithm in the multilayer network model, a new (MLCD) algorithm for community detection in multi-layer networks is proposed. The algorithm can be applied to complex network models of heterogeneous relationships. Finally, the LFR framework of community performance detection is used, and the results of MLCD are compared with those of the popular Louvain and Infomap community discovery algorithms, and the applicability and effectiveness of this algorithm are confirmed. 2. In this paper, an algorithm for extracting the core influence structure of complex networks is proposed. In this algorithm, the core motifs in the neighborhood subgraph of each node in the network are mined, and then combined to form the core influence structure. Different from the traditional core structure mining method, the core influence structure is a concise network subgraph, which not only includes the core nodes in the network, but also describes the relationship between the core nodes and the non-core nodes. At the same time, the structure can well reflect the topology and scale characteristics of the original network. This method is suitable for network parameter estimation, visual analysis and so on. It can also be used to extract the network topology of complex systems. To sum up, this paper takes the edge pattern as the basic object of the network, studies the relationship heterogeneity and the core influence of the complex network deeply, and obtains the effective results. Therefore, the analysis method of complex network based on edge pattern can be used as an important research tool for the development of complex network in the future.
【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
【参考文献】
相关期刊论文 前1条
1 汪小帆;刘亚冰;;复杂网络中的社团结构算法综述[J];电子科技大学学报;2009年05期
相关硕士学位论文 前1条
1 任成磊;社会网络的邻域重叠社团划分[D];华东师范大学;2016年
，本文编号：2331562
本文链接：https://www.wllwen.com/kejilunwen/yysx/2331562.html
