多个动态序列之间非线性相关性度量方法

Measuring approach for nonlinear dependence among time series

动态序列相关性推断是处理多维序列经济预测建模中变量选择问题的重要方法之一.本文把状态空间重构意义下相关维数计算方法推广到任意多个观测变量情形,定义了多个序列非线性相关度的概念,并讨论了计算方法的量纲稳定性.利用Lorenz系统进行仿真,并以中房综合指数上海、北京、广州数据作为应用实例进行非线性相关性推断,数据结果说明方法可行.

Dependence inference between dynamics series is one of the most important approaches for variable selection in multivariate time series economic forecasting modeling.In this paper,the reconstruction principle and correlation dimension estimation method are extended to the case of multivariate observations,at the same time the nonlinear dependence coefficient of multivariate series is defined,and the dimensional stability of estimation method is discussed.The simulation result by using Lorenz system,and nonl...