基于前兆数据和经验模态分解的金融系统极值风险识别研究  

Research on the Extreme Risk Identification of Financial System in Basis of Precursory Data and Empirical Mode Decomposition

作　　者：郭建平 赵立龙 GUO Jian-ping;ZHAO Li-long(School of Management Science and Engineering,Nanjing University of Information Science&Technology,Nanjing 210044,China;School of Physics&Optoelectronic Engineer,Nanjing University of Information Science&Technology,Nanjing 210044,China)

机构地区：[1]南京信息工程大学管理工程学院,江苏南京210044 [2]南京信息工程大学物理与光电工程学院,江苏南京210044

出　　处：《统计与信息论坛》2021年第3期60-69,共10页Journal of Statistics and Information

基　　金：教育部人文社会科学研究规划基金项目“频域分析视角下金融系统极值风险防范与化解机制研究”(19YJA 630023)。

摘　　要：基于证券市场综合指数,通过分割样本数据形成系列子样本,构造F统计量。研究不同子样本之间的差异,识别前兆数据样本,采用经验模态分解算法将前兆数据分解成不同频率成分,分析各频率成分变化特征,识别极值生成的重要特征,对前兆数据样本中的极值进行识别研究。结果表明,不同频率模态序列的波峰叠加特征和振幅持续增大特征的同时存在是极值生成的重要特征。若波峰叠加特征与振幅持续降低特征同时存在,则生成极值的可能性大大降低,极值大小由当时随机冲击的强度决定。并且,系统随机冲击因素并非极值生成的决定性条件,没有随机冲击因素,系统也可生成极值。研究结果可为有效防范和化解金融系统的极值风险提供可操作性的参考依据。Based on the comprehensive index of the stock market,a series of sub samples are formed by dividing the sample data,and F-statistic are constructed.The extreme values are identified in the precursor data samples by studding the differences between different sub samples,identifying the precursor data samples,using empirical mode decomposition algorithm to decompose the precursor data into different frequency components,analyzing the change characteristics of each frequency component,identifying the important features of extreme value generation.The results show that the important feature of extreme value generation is the simultaneous existence of peak superposition feature and amplitude increasing feature of different frequency intrinsic mode sequences.If the superposition of wave peaks and the continuous decrease of amplitude exist at the same time,the possibility of generating extreme value is greatly reduced,and the magnitude of extreme value is determined by the strength of random impact at that time.Moreover,the random impact factors of the system are not the decisive conditions for the generation of extreme value.Without random impact factors,the system can also generate extreme value.The study provides a reference for preventing and resolving the extreme risk of financial system.

关 键 词：前兆数据 F统计量 经验模态分解 极值识别 

分 类 号：C812[社会学—统计学]