Neural Network Prediction of Self-Similarity Network Traffic

Abstract

Several factors are found to influence either short or long-term
burstiness in Transmission Control Protocol (TCP) flow across many
networking facilities and services. Predicting such self-similar traffic
has become necessary to achieve better performance. In this study,
ANN model was deployed to simulate College Campus network traffic.
A Feed Forward Backpropagation Artificial Neural Network (ANN)
and Wireshark tools were implemented to study the network Scenario.
The predicted series were then compared with the corresponding real
traffic series (Mobile Telephone-Network (MTN)-Nigeria). Suitable
performance measurements of the Means Square Error (MSE) and the
Regression Coefficient were used. Our results showed that burstiness
is present in the network across many time scales. With the increasing
number of data packet distributions thereby providing a steady flow
of burst over the entire period of system load as the traffic network
performance improves.