基于自相关函数的非平稳时序数据的辨识改进

Improved identification of non-stationary time series data based on autocorrelation function

由于自相关函数刻画了时序数据在不同时刻取值的线性相关程度,故其在时序数据的统计分析中得到了广泛的应用。讨论了基于FFT变换的自相关函数计算原理,结合非平稳时序数据的辨识需求,基于自相关函数理论对趋势和周期成份的分离次序以及残留序列的随机类型识别等问题进行了深入分析,进一步提出了一种改进的非平稳时序数据的辨识算法。实验验证了改进算法的合理性和有效性。

Because the autocorrelation function is used to describe the linear dependence of the time series data at different time,it has been widely used in the statistical analysis of the time series data.This paper discussed the calculation principle of FFT transform autocorrelation function,combined with identification requirements of non stationary time series data,based on autocorrelation function theory,analysed the trend and cycle component data separation order and residual sequence of random type recognition problem indepth.Furthermore,an improved non stationary time series data identification algorithm is proposed.The rationality and validity of the improved algorithm are verified by experiments.