Fast optimization method: an on-line hurst parameter estimator

Abstract

An extended version of optimization method (OM) for on-line Hurst estimation is presented named as fast optimization method (FOM). The on-line Hurst estimator is crucial to characterize self-similar feature on stochastic process and widely applied in various fields such as in network traffic analysis, bandwidth provisioning and anomaly detection. Recent on-line Hurst estimator based on fast wavelet transform known as real-time wavelet estimator (RWM) is proven can estimates faster than other methods in on-line fashion. However this paper will present the capability of FOM to estimates the Hurst parameter faster than RWM with an acceptable Hurst value, enabling it's to be used in on-line application. In order to verify FOM result and its performance, the method was implemented using two types of self-similar processes that are fractional Gaussian noise (fGn) and MIT/DARPA network traffic data set. The results show a significant improvement on FOM performance compared with RWM. While the estimated Hurst parameter locates in the range of confidence interval of OM method had proven the accuracy of FOM estimation.