邻域重叠社团结构的动态证券网络研究

发布时间：2018-06-25 02:09

本文选题：邻域重叠 + 完全子图　；参考：《华东师范大学》2012年硕士论文

【摘要】：在社会网络、神经网络、万维网络、通信网络等中,存在着一种特性——社团结构。所以,复杂网络社团结构探测研究,对于拓扑结构分析、功能的理解、网络动力学行为预测等有着重要的意义。是众多科学家研究兴趣所在。社团结构的划分可分为邻域重叠社团划分和非邻域重叠社团划分。现在许多邻域重叠社团结构探测算法能够快速而准确地实现中小规模的网络划分,但是网络中存在一些节点,它们完全映射,即两两相连,如果这种结构的团簇(cluster)数量增加,节点和边的可见度下降,许多算法不能兼容。基于上述问题,本文提出了基于完全子图(clique)的邻域重叠社团结构探测算法,针对具有社团结构的加权网络模型进行研究。以SP500和CSI300两个具有重叠结构的证券网络为研究对象,利用邻域重叠社团结构探测算法,去研究证券市场的特性,如网络拓扑结构、聚类特性、重叠节点等。论文主要贡献如下： 1.在进行社团划分时,划分步骤为：(1)寻找网络中clqiues作为社团核心。(2)合并完全子图。(3)划分非clique节点。本文对合并完全子图步骤进行改进,提出利用重叠程度参数指标合并完全子图,再利用cliques之间的连接度指标进一步合并的理论。在处理“非clique节点”时,提出四种划分方法,寻找最优划分达到最佳划分效果。 2.对沪深300(CSI300)和SP500在金融危机前和金融危机下进行社团的划分,寻找重叠节点。通过比较在金融危机冲击下的新兴市场(CSI300)和成熟市场(S＆P500),他们的共同点是：社团数目明显下降。说明随着全球化进一步加深,两个市场行业格局不再明显。同时,也发现了一些不同之处：美国市场的金融大震荡,使得社团结构完全破坏,而中国市场还有一定的社团结构。说明两市场受到金融危机冲击程度的不同。结合实际,发现CSI300海运行业、全国煤炭消费行业在金融危机下受到了非常大的冲击。SP500中医药行业在金融危机的冲击下多数比较乐观。 3.以CSI300从2005年1月到2010年12月的股票收盘价为研究对象,通过分析社团结构的时间演化趋势和社团之间重叠节点的变化规律。发现：社团结构从清晰到模糊,再到清晰的分布特性。CSI300在2005年没有重叠节点,随着时间的推移,重叠节点的个数呈递增的趋势。在2008年,重叠节点个数达到了一个峰值。随后,节点个数呈下降的趋势。本论文研究得到了国家自然科学基金项目的支持,项目名称《开放式空间加权网络的拓扑演化及其优化设计》,项目编号：11075057。
[Abstract]:In the social network, the neural network, the world wide web and the communication network, there is a characteristic of the community structure. Therefore, the research of the complex network community structure detection is of great significance for the topology analysis, the understanding of the function and the prediction of the network dynamic behavior. It is the interest of many scientists.
The division of community structure can be divided into neighborhood overlapping community division and non neighborhood overlapping community division. Now many neighborhood overlapping community structure detection algorithms can quickly and accurately realize small and medium scale network partition, but there are some nodes in the network, they are fully mapped, that is, 22 connected, if the number of clusters (cluster) of this structure is number. As the volume increases, visibility of nodes and edges decreases, and many algorithms are not compatible.
Based on the above problems, this paper proposes a neighborhood overlapping community structure detection algorithm based on complete subgraph (clique), and studies the weighted network model with community structure. Two overlapping security networks with SP500 and CSI300 are taken as the research object, and the neighborhood reiterative association structure detection algorithm is used to study the securities market. Characteristics such as network topology, clustering characteristics, overlapping nodes and so on.
The main contributions of the paper are as follows:
1. in the division of community, the steps are as follows: (1) looking for clqiues in the network as the core of the community. (2) merging complete subgraphs. (3) dividing the non clique nodes. This paper improves the steps of the combined complete subgraph, and proposes the combination of the overlap degree parameter index to the complete subgraph, and then use the connectivity index between cliques to further merge. In dealing with the "non clique node", four methods are proposed to find the optimal partition to achieve the best division effect.
2. to Shanghai, Shenzhen 300 (CSI300) and SP500 in the division of societies before and under the financial crisis, looking for overlapping nodes. By comparing the emerging markets (CSI300) and mature markets (S & P 500) under the impact of the financial crisis, their common point is that the number of societies has fallen significantly. With the further deepening of globalization, the two market sectors have been further deepened. The situation is no longer obvious. At the same time, some differences are found: the financial turmoil in the United States market has made the community structure completely destroyed, and the Chinese market has a certain community structure. It shows that the two market is affected by the financial crisis. In combination with the reality, the CSI300 shipping industry, the national coal consumption industry is subjected to the financial crisis. To a very big impact,.SP500 Chinese medicine industry under the impact of the financial crisis, most of the more optimistic.
3. the stock closing price of CSI300 from January 2005 to December 2010 was studied. By analyzing the time evolution trend of the community structure and the changing rules of overlapping nodes between societies, it was found that the organization structure from clear to fuzzy and then to clear distribution characteristics.CSI300 did not have overlapping nodes in 2005, and overlapped nodes over time. The number of overlapped nodes reached a peak in 2008, and the number of nodes decreased.
The research is supported by the National Natural Science Foundation project. The project name, the topology evolution of the open space weighted network and its optimization design, project number: 11075057.
【学位授予单位】：华东师范大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F832.51;F224

【相似文献】