摘要：Recent studies have shown that conventional traffic models are unable to capture the real
nature of traffic pattern in high-speed networks, such as ATM Networks. Instead, they found
that self-similar model is a better model in representing this type of traffic pattern. The
differences between conventional and self-similar models have significant implications in
network design, management and analysis. Due to these significant differences, researches
that suggest the implementation of neural network in traffic prediction could perform an
effective and accurate forecast, should be reviewed by using this new traffic model. With
this in mind, the use of neural network in predicting self-similar traffic in ATM Networks is
observed in this research. Neural Network is used because it has been proven to have very
good mapping abilities in some difficult tasks. In this paper, a (3-8-1) backpropagation
neural network is trained with a suitable sample and then used to predict self-similar traffic
in ATM networks. The results show that the trained neural network could effectively and
accurately predict the self-similar traffics in ATM Networks.
