Controlling the diversity in classifier ensembles through a measure of agreement

被引：22

作者：

Zouari, H ^{[1
]}

Heutte, L ^{[1
]}

Lecourtier, Y ^{[1
]}

机构：

[1] Univ Rouen, Lab PSI, FRE 2645, CNRS, F-76821 Mont St Aignan, France

来源：

PATTERN RECOGNITION | 2005年 / 38卷 / 11期

关键词：

diversity measure; classifier ensemble; output generation algorithm; classifier agreement; kappa measure; dependency; classifier simulation;

D O I：

10.1016/j.patcog.2005.02.012

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

In this paper, a simulation method is proposed to generate a set of classifier outputs with specified individual accuracies and fixed pairwise agreement. A diversity measure (kappa) is used to control the agreement among classifiers for building the classifier teams. The generated team outputs can be used to study the behaviour of class-type combination methods such as voting rules over multiple dependent classifiers. (c) 2004 Pattern Recognition Society. Published by Elsevier Ltd. All rights reserved.

引用

页码：2195 / 2199

页数：5

共 5 条

[1]

Fleiss J. L., 1981, Statistical Methods for Rates and Proportions, V2nd

[2] Generating classifier outputs of fixed accuracy and diversity [J].

Kuncheva, LI ;

Kountchev, RK .

PATTERN RECOGNITION LETTERS, 2002, 23 (05) :593-600

[3] Measures of diversity in classifier ensembles and their relationship with the ensemble accuracy [J].

Kuncheva, LI ;

Whitaker, CJ .

MACHINE LEARNING, 2003, 51 (02) :181-207

[4]

LECCE VD, 2000, 7 INT WORKSH FRONT H, P143

[5]

Zouari H, 2003, LECT NOTES COMPUT SC, V2709, P296

← 1 →