A dynamic classifier ensemble selection approach for noise data

被引:94
作者
Xiao, Jin [1 ]
He, Changzheng [1 ]
Jiang, Xiaoyi [2 ]
Liu, Dunhu [3 ]
机构
[1] Sichuan Univ, Sch Business Adm, Chengdu 610064, Sichuan Prov, Peoples R China
[2] Univ Munster, Dept Math & Comp Sci, D-48149 Munster, Germany
[3] Chengdu Univ Informat Technol, Fac Management, Chengdu 610103, Sichuan Prov, Peoples R China
关键词
Multiple classifier systems; GMDH; Dynamic ensemble selection; Noise-immunity ability; NEURAL-NETWORKS; COMBINATION; DIVERSITY; REGION;
D O I
10.1016/j.ins.2010.05.021
中图分类号
TP [自动化技术、计算机技术];
学科分类号
080201 [机械制造及其自动化];
摘要
Dynamic classifier ensemble selection (DCES) plays a strategic role in the field of multiple classifier systems. The real data to be classified often include a large amount of noise, so it is important to study the noise-immunity ability of various DCES strategies. This paper introduces a group method of data handling (GMDH) to DCES, and proposes a novel dynamic classifier ensemble selection strategy GDES-AD. It considers both accuracy and diversity in the process of ensemble selection. We experimentally test GDES-AD and six other ensemble strategies over 30 UCI data sets in three cases: the data sets do not include artificial noise, include class noise, and include attribute noise. Statistical analysis results show that GDES-AD has stronger noise-immunity ability than other strategies. In addition, we find out that Random Subspace is more suitable for GDES-AD compared with Bagging. Further, the bias-variance decomposition experiments for the classification errors of various strategies show that the stronger noise-immunity ability of GOES-AD is mainly due to the fact that it can reduce the bias in classification error better. (C) 2010 Elsevier Inc. All rights reserved.
引用
收藏
页码:3402 / 3421
页数:20
相关论文
共 64 条
[1]
[Anonymous], INF FUSION
[2]
[Anonymous], 1993, Proceedings of the 13th International Joint Conference on Artificial Intelligence
[3]
[Anonymous], 1963, THESIS PRINCETON U N
[4]
Banfield RE, 2003, LECT NOTES COMPUT SC, V2709, P306
[5]
A forecasting solution to the oil spill problem based on a hybrid intelligent system [J].
Baruque, Bruno ;
Corchado, Emilio ;
Mata, Aitor ;
Corchado, Juan M. .
INFORMATION SCIENCES, 2010, 180 (10) :2029-2043
[6]
Random forests [J].
Breiman, L .
MACHINE LEARNING, 2001, 45 (01) :5-32
[7]
Pattern recognition with neural networks combined by genetic algorithm [J].
Cho, SB .
FUZZY SETS AND SYSTEMS, 1999, 103 (02) :339-347
[8]
COMPOSITE CLASSIFIER SYSTEM-DESIGN - CONCEPTS AND METHODOLOGY [J].
DASARATHY, BV ;
SHEELA, BV .
PROCEEDINGS OF THE IEEE, 1979, 67 (05) :708-713
[9]
Demsar J, 2006, J MACH LEARN RES, V7, P1
[10]
A study on the performances of dynamic classifier selection based on local accuracy estimation [J].
Didaci, L ;
Giacinto, G ;
Roli, F ;
Marcialis, GL .
PATTERN RECOGNITION, 2005, 38 (11) :2188-2191