Estimation Techniques for Monitoring and Controlling the Performance of the Computer Communication Networks

Abstract

This study was concerned with making a comparative study between four types of time series model with the goal of choosing the optimum one to predict the performance of a computer communication network. The investigated four types of time series are: Least Square (LS), Fourier series, Exponential Weighted Moving Average (EWMA) and the Auto Regressive Integrated Moving Average (ARIMA). Comparative study is based on comparing some of statistical measurement for these four time series models as: mean standard deviation and variance. We apply these various types of time series on two types of network to predict their performance; the first one is called token bus while the second one is called the token ring. Concluded results prove that both EWMA and ARIMA perform better than the others LS and Fourier since both EWMA and ARIMA forecast the network performance parameters with an accuracy of 98% near the actual values.