股票市场相空间重构嵌入维与时滞的确定

Determinating the Embedding Dimension and Delay Time in Phase Space Reconstruction in Stock Market

现有方法在计算嵌入维m与时滞τ时都需要对m或τ等参数范围作合理的假定,而对于股票市场等复杂系统很难准确预知其有关非线性特征.针对这个问题,在C-C方法的基础上,首先讨论了m与τ等参数对关联积分的影响,然后利用τw=(m-1)τ为常数这一特性,通过对m进行迭代得到更优的嵌入维m与时滞τ,从而为进一步分析与预测股票市场提供了必要条件.

As the current method needs to make a reasonable assumption in the calculation of the embedding dimension mand time delay τ,it is too hard to accurately predict the nonlinear feature for the stock market and other complex system. Based on the C - C method, this paper first discusses the parameters impact on the correlation integral. Then, using τw = （ m - 1 ） τ as, a constant feature, we get a better embedding dimension m and time delay τ through iterationm, thus it provides the necessary condition for the further analysis and forecast in the stock market.