FARIMA model Hurst parameter estimation method based on domain searching

Abstract

The invention relates to an FARIMA model Hurst parameter estimation method based on domain searching. An FARIMA time sequence of the self-similarity long-range dependency characteristic possessed by simulating the actual network flow is used as an estimation source, and a mode of combining a time-variance graph method and a searching method is adopted for carrying out accurate estimation of Hurst parameters. The method specifically comprises the following steps that 1), the FARIMA time sequence of the network flow which can be simulated and has the self-similarity long-range dependency characteristic is generated; 2), an experience section used for roughly estimating a time block of the variance-time graph method of the FARIMA sequence is determined; 3), the variance-time graph method is used for carrying out Hurst parameter estimation of the FARIMA time sequence in the experience section obtained in the step 2); 4), the step size and precision parameters of the search method are determined, and precise search estimation of the Hurst parameters is carried out close to a rough estimation value. Compared with the prior art, the precision of Hurst parameter estimation is greatly improved.

Description

Technical field [0001] The invention relates to a wireless self-organizing network traffic prediction, in particular to a method for estimating Hurst parameters of a FARIMA model based on field search. Background technique [0002] As more and more studies have found that network traffic has the characteristic of self-similar long correlation, the research of this characteristic poses challenges to the modeling and prediction of network traffic. The traditional correlation model has large deviations, so it is more suitable to find and study. The model has important significance. After comparing several common network models in terms of performance, complexity and usage scenarios, it is believed that the FARIMA model has the characteristics of describing short- and long-correlation characteristics at the same time, and is particularly suitable for modeling and forecasting self-similar network traffic. Hurst parameters can be used to describe the self-similar long correlation char...

AI Technical Summary

Problems solved by technology

There are also many methods for estimating Hurst parameters, such as R/S method, wavelet method, and variance-time diagram method. Experiments have proved that these methods generally have low accuracy, and the choice of wavelet base in wavelet method

Images