A State Recognition Of System Dynamics Based On Moving Sample Entropy And Their Applications

Sample entropy can quantitatively describe dynamic state changes and complexity of deterministic and stochastic sequence,which has been used in biological medicine,fault diagnosis of bearing,signal processing,brain wave recognition and other fields.Firstly,this paper research on the applicability of M-SamEnt to identify dynamic state of linear and chaotic time series.Secondly,we discuss the effects of the parameters in the algorithm,sample size and noise interference on M-SamEnt in practice.At last,the application of the algorithm proposed to Shanghai stock index market complexity and state identification of geological metallogenic elements to provide new approaches for studying intrinsic nonlinear characteristics of financial time series and extracting dynamics feature of spatial geological data.The main research results are as follow:(1)Compared with the traditional method,slide t test and Mann-Kendall method,M-SamEnt method not only can identify dynamic state of linear and chaotic time series effectively but gives clear evolution curve and the result is more detailed and reliable.(2)Analysis of window size,sliding step,sample size,noise,influence on M-SamEnt method systematically.The result shows that the effects of window size,sliding size on mutation and dynamic state of linear and chaotic time series were minor.By using comparative analysis,performance of this method is more advantage than sliding t test and Mann-Kendall method for analyzing small sample sequences.In addition,sample entropy curve of linear and chaotic time series which added different intensity of Gaussian noise or White noise are different.The anti-noise capability of linear time series is better than chaotic time series and discover that the increment of SNR to span,recognition efficiency begins to stabilize.(3)By analyzing characteristics of dynamic states of the Shanghai Composite Index,providing a new approach idea for market regulators to analyze the volatility of share price,M-SamEnt can not only extract the time period of financial crisis greatly,but reflect complexity of stock market.(4)By analyzing characteristics of non-linearity of the elements Au from Shangzhuang Gold Mine in Shandong and the elements Cu from Pulang in Yunan with M-SamEnt,value of sample entropy of different content ore-forming elements can be an decision fundament of mineralization identification and can be used to quantify statistical characteristic of mineral deposit,and give references for mining potential mineral-forming areas.