作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程

• 先进计算与数据处理 • 上一篇    下一篇

基于多标度曲线的股市网络构造及其社区挖掘

袁 铭   

  1. (天津财经大学统计系,天津300222)
  • 收稿日期:2014-05-26 出版日期:2015-04-15 发布日期:2015-04-15
  • 作者简介:袁 铭(1982 - ),男,讲师、博士,主研方向:复杂网络,模式识别,金融时间序列分析。
  • 基金资助:
    天津市哲学社会科学规划基金资助项目(TJTJ13-002)。

Construction of Stock Market Network Based on Multi-scale Curve and Its Community Mining

YUAN Ming   

  1. (Department of Statistics,Tianjin University of Finance and Economic,Tianjin 300222,China)
  • Received:2014-05-26 Online:2015-04-15 Published:2015-04-15

摘要: 针对股票市场的复杂网络建模问题,提出使用不同阶数下的标度曲线(多标度曲线),测度沪深300 指标股之间的加权多重分形特征相似性,并据此构造网络,研究网络的拓扑性质。在此基础上采用快速Newman, Girvan-Newman,Louvain 等经典算法挖掘网络社区结构,利用最大模块度确定最优相似性门限值,通过投资组合MV 模型验证方法的有效性。实验结果表明,多标度曲线网络具有无标度、小世界和富人俱乐部性质,使用不同算 法挖掘其社区结构可得到最优的划分效果。基于该网络社区结构构造的投资组合可有效降低风险。

关键词: 多标度曲线, 复杂网络, 拓扑结构, 社区结构, 模块度, 投资组合MV 模型

Abstract: This paper proposes a method of building stock market network based on measuring weighted multi-fractal similarity among CSI 300 index stocks through multi-scale curve. It also studies the topology properties of the network. It tries to employ several classic algorithms,such as Fast Newman,Girvan-Newman,Louvain to find its community structure and determine the optimal threshold according to maximum modularity. This paper adopts portfolio MV model to further verify the method’s effectiveness. Experimental results show that the network based on multi-scale curve has scale-free, small world and rich club properties. This network has the best classification when using different community mining algorithms and displays robustness to these algorithms. The portfolio based on the community structure can reduce risk considerably.

Key words: multi-scale curve, complex network, topology structure, community structure, modularity, portfolio MV model

中图分类号: