基于样本熵的时间序列非线性检测方法研究

Nonlinear detection study of time series based on sample entropy

为检测时间序列的非线性特性,针对替代数据法中常用特征量的不足,提出一种基于样本熵的非线性检测方法。采用样本熵作为替代数据法中的特征量来检测时间序列的非线性,在Lorenz方程、Logistic方程以及线性AR模型产生的3种仿真数据上进行验证,进一步与其它算法的时间效率进行对比。实验结果表明,对于不同长度、不同特性的数据,该方法的检测结果稳定有效,时间效率大幅度提高。

To detect the non-linear characteristics of the time series, aiming at deficiencies of the common test statistics of the surrogate method, the nonlinear detection method based on the sample entropy was proposed. The sample entropy was used as the test statistic of surrogate method to detect the nonlinear character of the time series, and the method was verified on three kinds of simulation data that generated lay using the Logistic equation, the Lorenz equation and the linear AR model, the time ef- ficiency was further compared with other algorithms. The study shows that for the data with different characteristics and diffe- rent lengths, the detection results of the method are stable and effective, and the time efficiency is greatly improved.