多观测变量状态空间重构技术及应用

Reconstruction principle for multivariate time series and its application

为适应于多个观测变量的状态空间重构问题,对混沌与非线性经济学中状态空间重构技术与Takens嵌入定理进行了推广,给出了多个观测变量状态空间重构意义下相关维数的计算方法.作为应用讨论了度量两列时间序列非线性相关性的一种方法.仿真结果与应用实例说明,该方法可较好地用于解决非线性经济预测中的变量选择问题.

The reconstruction principle and Takens embedding theorem were generalized to multivariate time series. The estimator of correlation dimension based on multivariate time series was constructed. As its application example, an inference method was designed for testing nonlinear dependence between time series, which can be used in economic forecasting. Numerical results show the effectiveness of the method proposed.