多变量时间序列相空间重构中参数的确定

Determination of parameters in the phase-space reconstruction of multivariate time series

介绍了多变量时间序列相空间重构理论.提出一种新的基于平均预测误差最小化的重构参数确定方法,阐述了该方法的算法过程及一些重要特点.此方法考虑了所有重构参数对平均预测误差的影响,能够同时确定重构系统相空间所需的恰当嵌入维数及时间延迟.最后将该方法应用于股票市场非线性动力系统的相空间重构,通过比较和分析验证了其优越性.

Phase-space reconstruction theory of multivariate time series is introduced. A novel method of determining reconstruction parameters based on the minimization of average forecasting error is proposed, and its algorithm procedure and some important characteristics are explicated. The method considers the effects of all reconstruction parameters on average forecasting error. It can simultaneously determine the correct embedding dimensions and time delays needed for better reconstructing systematic phase-space. The method is applied to the phase-space reconstruction of nonlinear dynamical system of stock market. Through comparing and analyzing, its superiority is verified.