基于经验模式分解的扩散熵分析法

Diffusion entropy analysis based on empirical mode decomposition

由于现实中的许多时间序列含有较强的周期趋势,周期趋势通常会影响序列的相关性。为了正确分析序列的相关性,提出了基于经验模式分解的扩散熵分析法,并用来分析北京市交通流序列的长程相关性。数值实验结果表明:经验模式分解法能有效地去除时间序列的周期趋势;扩散熵分析法能求得去除周期趋势后的序列的正确的标度指数;北京交通流序列呈现长程相关性。

Strong periodic trend in many real time series often has effect on the correlation of series. In order to analyze the correlation of series correctly, diffusion entropy analysis （DEA） is presented based on Renyi entropy and the long correlation of Beijing traffic flow series is analyzed. The numerical experiments show： EMD can eliminate the periodic trend of series; DEA can get the correct scaling exponent after eliminating the periodic trend of series; Beijing traffic flow series show a long- range correlation.