Number of the records: 1

Dynamic Classifier Systems and their Applications to Random Forest Ensembles

SYSNO ASEP	0326646
Document Type	C - Proceedings Paper (int. conf.)
R&D Document Type	Conference Paper
Title	Dynamic Classifier Systems and their Applications to Random Forest Ensembles
Title	Dynamické systémy klasifikátorů a jejich využití pro klasifikaci pomocí náhodných lesů
Author(s)	Štefka, David (UIVT-O) Holeňa, Martin (UIVT-O)_{SAI, RID}
Source Title	Adaptive and Natural Computing Algorithms. - Berlin : Springer, 2009 / Kolehmainen M. ; Toivanen P. ; Beliczynski B. - ISBN 978-3-642-04920-0
Pages	s. 458-468
Number of pages	11 s.
Action	ICANNGA'2009. International conference /9./
Event date	23.04.2009-25.04.2009
VEvent location	Kuopio
Country	FI - Finland
Event type	WRD
Language	eng - English
Country	DE - Germany
Keywords	classifier combining ; dynamic classifier aggregation ; random forests ; classification
Subject RIV	IN - Informatics, Computer Science
R&D Projects	1ET100300517 GA AV ČR - Academy of Sciences of the Czech Republic (AV ČR)
	GA201/08/0802 GA ČR - Czech Science Foundation (CSF)
CEZ	AV0Z10300504 - UIVT-O (2005-2011)
UT WOS	000279120700047
EID SCOPUS	78650738728
DOI	10.1007/978-3-642-04921-7_47
Annotation	Classifier combining is a popular method for improving quality of classification -- instead of using one classifier, several classifiers are organized into a classifier system and their results are aggregated into a final prediction. However, most of the commonly used aggregation methods are static, i.e., they do not adapt to the currently classified pattern. In this paper, we provide a general framework for dynamic classifier systems, which use dynamic confidence measures to adapt to a particular pattern. Our experiments with random forests on 5 artificial and 11 real-world benchmark datasets show that dynamic classifier systems can significantly outperform both confidence-free and static classifier systems.
Workplace	Institute of Computer Science
Contact	Tereza Šírová, sirova@cs.cas.cz, Tel.: 266 053 800
Year of Publishing	2010

Number of the records: 1