7种Hurst系数估计算法的性能分析

讨论影响Hurst系数估计算法的因素，包括方差、周期信号和相关结构。调整分形高斯噪声（FGN）序列，从而产生具有部分尺度范围相关结构的序列。不断改变尺度范围并估计序列的Hurst系数，发现算法的估计结果依赖于特定尺度范围的相关结构，而尺度范围以外的相关结构的改变对估计结果无影响。对于实际业务流量，相关结构的变化导致算法估计结果的不同。

Performance analysis of seven estimate algorithms about the Hurst coefficient

The factors which affect the performance of estimate algorithms about Hurst coefficient was discussed,including variance, periodic signal and correlation structure. A new serial of correlation structure in a specific scale range was constructed by rearranging FGN(Fractal Gaussian Noise) serial, the scale range was changed continuously and the new serial was estimated. The estimation of each algorithm depended on the correlation structure in a specific scale range, but the structures out of the scale range had no effects on the estimation. For the practical network traffic, the estimations of the algorithms vary according to the correlation structures.