复杂性度量方法的改进及其在证券市场的应用

Improvement of complexity measure method and its application to the stock exchange market

系统地分析了Lemple-Ziv复杂性度量方法的应用过程中,将实际信号(时间序列)转变成符号序列的诸多方法中存在的一些问题,提出了更合理的处理方法,即兼容法.兼容法可以有效地刻画各种时间序列的复杂度,克服了均值法等方法中将序列过分"粗粒化"以致不能区分混沌与完全随机现象的缺点,克服了极值法与遗传密码粗粒化方法因只提取细节成分信息而导致辨别率不高等缺点.最后动态地分析了中国证券市场的复杂性.

The paper analyzes the problems of transforming signals （i. e. time series） into symbol series in the application of the Lemple-Ziv theory of complexity and then proposes a more suitable method, namely the compatible method. The method can not only describe the complexity of different time series, but also avoid the weakness of oversimplified coarse grain of state space that will result in that chaos and complete random series can not be differentiated in the mean methocl, and the weaknesses of only paying attentionto local information that will result in ineffectiveness of distinguishing time series in the extremal method and the genetic code method. Finally the dynamic complexity of the Shanghai Stock Exchange Market is analyzed.