基于连接密度的网络社团发现方法研究与实现

发布时间：2018-06-24 03:23

本文选题：社团发现 + 局部社团　；参考：《西安电子科技大学》2012年硕士论文

【摘要】：伴随着互联网及相关信息技术的迅猛发展，人类社会已经迈入了“网络时代”。复杂网络（Complex Network）系统随处可见。复杂网络的一个重要属性是其内在的社团结构。挖掘网络的社团结构，有助于揭示网络拓扑结构，分析其功能和隐含模式，预测复杂网络的行为，在社会网络、生物信息、搜索引擎、信息推荐、精准营销等领域都有较大的应用价值。本文首先深入研究了网络社团的定义、结构及当前主流的网络社团发现方法，分析其优缺点。基于此提出了一种基于节点连接密度的局部社团质量评价标准，并据此提出一个快速在线多分辨率重叠社团检测算法。该算法无需网络全局信息和用户干预，能够从网络中的任一节点开始，自动检测包含起始节点的局部社团。其次，本文将基于连接密度的网络聚类方法和基于最小生成树的网络聚类方法相结合，提出了一种基于网络骨架树的密度网络社团发现方法。该算法能准确的识别出各种规模的社团以及中心点和离群点,并能定位最优密度阈值。最后,在真实网络和生成网络上对算法进行了详细的对比实验。结果表明：本文提出的网络社团发现方法准确率高，运行速度快，能够处理各种规模和形状的网络。
[Abstract]:With the rapid development of Internet and related information technology, human society has entered the "network era". Complex network systems can be found everywhere. An important attribute of complex network is its internal community structure. Mining the community structure of the network helps to reveal the topology of the network, analyze its functions and hidden patterns, predict the behavior of complex networks, in social networks, biological information, search engines, information recommendation, Precision marketing and other fields have greater application value. In this paper, the definition and structure of network community and the current mainstream network community discovery methods are studied, and their advantages and disadvantages are analyzed. Based on this, a local community quality evaluation standard based on node connection density is proposed, and a fast online multi-resolution overlapping community detection algorithm is proposed. Without global information and user intervention, the algorithm can automatically detect local communities including starting nodes from any node in the network. Secondly, this paper combines the network clustering method based on connection density with the network clustering method based on minimum spanning tree, and proposes a density network community discovery method based on network skeleton tree. The algorithm can accurately identify the community, center and outliers of various sizes, and can locate the optimal density threshold. Finally, the algorithm is compared with real network and generating network in detail. The results show that the network community discovery method proposed in this paper has the advantages of high accuracy and fast running speed, and it can deal with networks of various sizes and shapes.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP301.6;O157.5

【相似文献】