基于非线性波动网络模型的股票市场关联特征研究  

Research on the correlation characteristics of stock market based on nonlinear volatility network

作　　者：李为波 郭雪 LI Weibo;GUO Xue(School of Economics,Wuhan Textile University, Wuhan 430200, China)

机构地区：[1]武汉纺织大学经济学院,湖北武汉430200

出　　处：《湖北大学学报（自然科学版）》2021年第2期192-199,共8页Journal of Hubei University：Natural Science

基　　金：教育部科技发展中心高校产业研创新基金-新一代信息技术创新项目资助课题“基于云计算与大数据的智慧教育云平台研究”项目(2018A02024)资助。

摘　　要：股价波动是股票市场研究的重要内容,本研究基于信息理论和复杂网络理论提出股票市场的股价波动网络模型.首先对股价波动关联性进行度量,然后分析股价波动的聚集性,运用极大平面过滤图算法对股票之间关联性进行分层研究.以上证180指数金融成分股票为样本,实证结果发现:与常用的线性相关系数相比,在非线性波动条件下,互信息和最大信息系数对关联系有更好的度量,在结果的精确性上,最大信息系数比互信息更具优势.在构建的股票网络中,仅有少数关键的股票节点,信息的传递通过这些节点以分层的方式传递到整个市场.在金融大数据背景下,非线性波动网络模型为挖掘市场风险特征提供新的方法.Stock price volatility is an important part of stock market.Based on information theory and complex network theory,we proposed a network model of stock market price volatility.Firstly,we measured the correlation of stock price volatility,and analyzed the aggregation of stock price volatility.Then we applied the planar maximally filtered graph algorithm to study the relationships between stocks.In this study,the empirical results show that:comparing with the commonly used linear correlation coefficient,mutual information and maximum information coefficient have better measurement of correlation under the condition of nonlinear fluctuation,and the maximum information coefficient has more advantages than mutual information in the accuracy of the results.In the stock network,there are only a few key stock nodes,and the information can be transmitted to the whole market in a hierarchical way through these nodes.In the context of financial big data,the nonlinear volatility network model provides a new method of mining market risk characteristics.

关 键 词：复杂网络 互信息 线性相关系数 最大信息系数 极大平面过滤图 

分 类 号：C812[社会学—统计学] O242[理学—计算数学]