A Neural Network Model for Improved Internet Service Resource Provisioning
M. O. Odim *
Department of Mathematical Sciences, Redeemer’s University, Nigeria.
J. A. Gbadeyan
Department of Mathematics, University of Ilorin, Nigeria.
J. S. Sadiku
Department of Computer Science, University of Ilorin, Nigeria.
*Author to whom correspondence should be addressed.
Abstract
The paper seeks for a good forecast model that can accurately represent the inherent characteristics of Internet traffic and forecast the desired traffic load to satisfy the performance target using Artificial Neural network technology. We developed a computational tool in Visual Basic 6.0 for this purpose, based on a Multi-layer network. By making use of an empirical study to examine the effect of some ANN model design issues such as impact of lag observation and the number of neurons of a two hidden layer network on Internet traffic prediction, the results shows that ANN is a powerful forecast modeling tool that can accurately capture the inherent traffic characteristics and forecast the desired traffic load and that these factors have greater impact on the performance ANN forecaster.
Keywords: Internet traffic, times series forecasting, multi-layer artificial neural network, resource provisioning, quality of service.