Welcome to the InfoSci Platform
Congestion Prediction System With Artificial Neural Networks

Congestion Prediction System With Artificial Neural Networks

Fatma Gumus, Derya Yiltas-Kaplan
Copyright: © 2020 |Volume: 12 |Issue: 3 |Pages: 16
ISSN: 1941-8663|EISSN: 1941-8671|EISBN13: 9781799806035|DOI: 10.4018/IJITN.2020070103
Cite Article Cite Article

MLA

Gumus, Fatma, and Derya Yiltas-Kaplan. "Congestion Prediction System With Artificial Neural Networks." IJITN vol.12, no.3 2020: pp.28-43. http://doi.org/10.4018/IJITN.2020070103

APA

Gumus, F. & Yiltas-Kaplan, D. (2020). Congestion Prediction System With Artificial Neural Networks. International Journal of Interdisciplinary Telecommunications and Networking (IJITN), 12(3), 28-43. http://doi.org/10.4018/IJITN.2020070103

Chicago

Gumus, Fatma, and Derya Yiltas-Kaplan. "Congestion Prediction System With Artificial Neural Networks," International Journal of Interdisciplinary Telecommunications and Networking (IJITN) 12, no.3: 28-43. http://doi.org/10.4018/IJITN.2020070103

Export Reference

Mendeley
Favorite Full-Issue Download

Abstract

Software Defined Network (SDN) is a programmable network architecture that provides innovative solutions to the problems of the traditional networks. Congestion control is still an uncharted territory for this technology. In this work, a congestion prediction scheme has been developed by using neural networks. Minimum Redundancy Maximum Relevance (mRMR) feature selection algorithm was performed on the data collected from the OMNET++ simulation. The novelty of this study also covers the implementation of mRMR in an SDN congestion prediction problem. After evaluating the relevance scores, two highest ranking features were used. On the learning stage Nonlinear Autoregressive Exogenous Neural Network (NARX), Nonlinear Autoregressive Neural Network, and Nonlinear Feedforward Neural Network algorithms were executed. These algorithms had not been used before in SDNs according to the best of the authors knowledge. The experiments represented that NARX was the best prediction algorithm. This machine learning approach can be easily integrated to different topologies and application areas.

Request Access

You do not own this content. Please login to recommend this title to your institution's librarian or purchase it from the IGI Global bookstore.