Network traffic classification with improved random forest

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Accurate network traffic classification is significant to numerous network activities, such as QoS and network management etc. While port-based or payload-based classification methods are becoming more and more difficult, Machine Learning methods are promising in many aspects. In this paper, we improve the standard Random Forest by setting the variable selection probability according to the importance of the corresponding variable to classify network traffic. Our test results show that the Improved Random Forest has better classification performance. And it takes less time to build the model.

Original languageEnglish
Title of host publicationProceedings - 2015 11th International Conference on Computational Intelligence and Security, CIS 2015
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages78-81
Number of pages4
ISBN (Electronic)9781467386609
DOIs
StatePublished - 1 Feb 2016
Event11th International Conference on Computational Intelligence and Security, CIS 2015 - Shenzhen, China
Duration: 19 Dec 201520 Dec 2015

Publication series

NameProceedings - 2015 11th International Conference on Computational Intelligence and Security, CIS 2015

Conference

Conference11th International Conference on Computational Intelligence and Security, CIS 2015
Country/TerritoryChina
CityShenzhen
Period19/12/1520/12/15

Keywords

  • Random forest
  • Selection probability
  • Traffic classification

Fingerprint

Dive into the research topics of 'Network traffic classification with improved random forest'. Together they form a unique fingerprint.

Cite this