Title: Analysis of prediction performance of training-based models using real network traffic

Authors: Mohamed Faten Zhani, Halima Elbiaze, Farouk Kamoun

Addresses: Department of Computer Science, University of Quebec in Montreal, Canada. ' Department of Computer Science, University of Quebec in Montreal, Canada. ' National School of Computer Sciences, Manouba 4010, Tunisia

Abstract: Traffic prediction constitutes a hot research topic of network metrology. This work focuses on the design and the empirical evaluation of the behaviour of training-based models for predicting the input rate of a single link. Via experimentation on real network traffic, we study the effect of some parameters on the prediction performance such as the amount of data needed to identify the model, the inputs of the model, the data granularity, and packet size distribution. Experiments show that the models provide accurate prediction using one lag. We also show that count of large packets is sufficient to predict the throughput.

Keywords: traffic measurements; traffic modelling; traffic prediction; neurofuzzy models; ARIMA model; self-similarity; neural networks; fuzzy logic; network traffic; short period throughput; training.

DOI: 10.1504/IJCAT.2010.030471

International Journal of Computer Applications in Technology, 2010 Vol.37 No.1, pp.10 - 19

Published online: 17 Dec 2009 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article