Simulation Modeling Network Traffic Behavior Using Regression Analysis in Wavelet Domain

This paper introduces a very simple method for building a simulation model of network traffic with noisy behavior in order to obtain more "readable" visualization. Initially, after the fast wavelet transform (FWT) of statistical data, it is suggested to replace the wavelet coefficients with their estimates received from the computed probability densities in all subbands of the FWT. Then, regression analysis is applied to all the modified subbands to obtain trends for both approximation and details. The curve of the traffic behavior is built after the inverse FWT when the possible negative values in the restored data are filtered. The results of modeling have shown that the obtained method outperforms some known regression methods in clarity and simplicity.