Application of moving cut data-rescaled variance analysis in dynamic structure mutation testing

Abstract

Scaling exponent is a valid nonlinear dynamics index, and can effectively detect the dynamic structure mutations in the correlation time series by means of moving a fixed widow and the moving cut a fixed window technique. When there is short-term correlation, sequence with short-term correlation and the calculation result of scaling exponent with rescaled range will be influenced to a certain degree, resulting in a certain deviation for the scaling exponent calculation of a moving cut window sequence, and some false mutations point and mutations range for abrupt change detection. In view of this, we present a new method of the dynamic structure mutation detection-moving cut data-rescaled variance analysis. The numerical testing of ideal time series shows that the moving cut data-rescaled variance analysis has strong stability and accuracy, which is much better than the moving rescaled variance and the moving approximate entropy. The test results have not false mutation point and interval when the moving window is small. The further application to practically measured data validates the reliability of the new method.