Stability Appraisal of the Wavelet-and-Regression Method for Unsteady Network Traffic Trend Highlighting

In this paper, we investigate the stability of the suggested method for trend highlighting from network traffic data based on a combination of wavelet transform and regression analysis. To obtain stability appraisal when a network traffic is unsteady, the initial trend highlighted from the real network traffic data was distorted by noise components using three models: the autoregressive (AR) model, the Markov modulated Poisson process (MMPP) model and the ON-OFF model. The results of re-highlighting the trend have shown that the proposed method allows preserving the features of network traffic while minimizing distortions caused by the noise components.