Abstract

Random Forest is a popular decision forest building algorithm which focuses on generating diverse deci-sion trees as the base classifiers. For high dimensional data sets, Random Forest generally excels in generat-ing diverse decision trees at the cost of less accurate individual decision trees. To achieve higher prediction accuracy, a decision forest needs both accurate and diverse decision trees as the base classifiers. In this paper we propose a novel decision forest algorithm called Complement Random Forest that aims to gen-erate accurate yet diverse decision trees when applied on high dimensional data sets. We conduct an elab-orate experimental analysis on seven publicly avail-able data sets from UCI Machine Learning Reposi-tory. The experimental results indicate the effective-ness of our proposed technique.
Original languageEnglish
Title of host publicationProceedings of the Thirteenth Australasian Data Mining Conference (AusDM 15)
Place of PublicationAustralia
PublisherCRPIT
Pages89-97
Number of pages9
Publication statusPublished - 2015
EventThe 13th Australasian Data Mining Conference: AusDM 2015 - University of Technology, Sydney, Australia
Duration: 08 Aug 201509 Aug 2015

Conference

ConferenceThe 13th Australasian Data Mining Conference
Country/TerritoryAustralia
CitySydney
Period08/08/1509/08/15

Fingerprint

Dive into the research topics of 'Complement random forest'. Together they form a unique fingerprint.

Cite this