ATM VOICE TRAFFIC . PREDICTION USING NEURAL NETWORKS

Abstract

Asynchronous Transfer Mode(ATM) Broad band networks support a range of multimedia traffic(e.g, voice, video, image, and data). Accurate characterization of the multimedia traffic is essential, in ATM networks, in order to develop a robust set of traffic descriptors. Such a set is required, for the important functions of traffic enforcement(policing) and bandwidth allocation. In this dissertation neural networks (NNs) are used for characterizing and predicting the voice traffic(from N number of sources). A backpropagation neural network is trained to predict the arrival process over a fixed measurement period of time, based upon sampled values taken from the previous measurement period. The performance of the standard algorithm and a modified version of it have been studied. Accuracy of traffic prediction has been verified by matching variance and index of dispersion for count process, to those of the ANN output. Performance of the two backpropagation algorithms is compared with the help of rate of convergence, in the light of a specific problem. It is found that the modified backpropagation converges in less number of iterations than standard backpropagation algorithm. The program is written in C & C + + under UNIX operating system.