A posteriori Fractal Characteristics Change Point Detecting Algorithm for Time Series

The construction of a posteriori fractal characteristics change point detecting algorithm for fractal time series is considered. It is proposed to use a geometrical fractal index to solve this problem, which has a significantly lower error in the estimation of the Hurst exponent compared to other known methods, even for small samples. The proposed algorithm based on the determining the intersection point for the two approximating lines for cumulative curvature graph. The efficiency of the algorithm is illustrated on the examples of the real time series.