Research On Time-varying Related Parameter Estimation Method And Its Application In Network Traffic Data Analysis

The traditional Hurst exponent is a constant,which can only describe the self-similarity of the entire data.The estimation results cannot describe the information of local mutation of these data.However,recent studies have shown that a large number of self-similar data reflect local self-similar characteristics.Compared to the self-similar parameters of the overall data,the local Hurst exponent is different.The local self-similarity of data can effectively describe the local relevant characteristics of the data and it is a generalized form of traditional self-similar parameters.Therefore,the analysis of local self-similar characteristics of data has gradually received attention.The local self-similarity of time series data is studied in this research.The establishment of more accurate system models and the prediction of future trends are of great significance.In order to analyze the local self-similarity of data and improve the accuracy of time-varying Hurst exponent estimation,the local catastrophe environment of time series simulation data with Alpha stable distribution noise is used.The traditional R/S estimation algorithm has the defects of low calculation accuracy and poor efficiency.In this research,the re-scaling part of its algorithm is improved,and an revised R/S estimation algorithm based on sequence length common divisor is provided.The reliability of time-varying Hurst exponent estimation in time series under noisy environment is analyzed using improved estimation algorithm combined with sliding window function.In addition,in order to verify the effectiveness of the time-varying Hurst exponent estimation algorithm in this study,an improved R/S estimation algorithm combined with sliding window function is applied to actual network traffic and get better analysis results.The improved algorithm given in this thesis eliminates the deficiency of the traditional R/S estimation algorithm in the rescaling method and expands the scope of application of the Hurst exponent estimation algorithm.This improved algorithm provides an effective method for self-similarity analysis of network traffic data.