A boosting-based framework for self-similar and non-linear internet traffic prediction

Hanghang Tong, Chongrong Li, Jingrui He

Research output: Contribution to journalArticle

5 Scopus citations

Abstract

Internet traffic prediction plays a fundamental role in network design, management, control, and optimization. The self-similar and non-linear nature of network traffic makes highly accurate prediction difficult. In this paper, a boosting-based framework is proposed for self-similar and non-linear traffic prediction by considering it as a classical regression problem. The framework is based on Ada-Boost on the whole. It adopts Principle Component Analysis as an optional step to take advantage of self-similar nature of traffic while avoiding the disadvantage of self-similarity. Feed-forward neural network is used as the basic regressor to capture the non-linear relationship within the traffic. Experimental results on real network traffic validate the effectiveness of the proposed framework.

Original languageEnglish (US)
Pages (from-to)931-936
Number of pages6
JournalLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume3174
StatePublished - Dec 1 2004

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint Dive into the research topics of 'A boosting-based framework for self-similar and non-linear internet traffic prediction'. Together they form a unique fingerprint.

  • Cite this