Classification by Cluster Analysis: A New Meta-Learning Based Approach

Research output: Chapter in Book/Report/Conference proceedingChapter

9 Citations (Scopus)

Abstract

Combination of multiple classifiers, commonly referred to as an classifier ensemble, has previously demonstrated the ability to improve classification accuracy in many application domains. One popular approach to building such a combination of classifiers is known as stacking and is based on a meta-learning approach. In this work we investigate a modified version of stacking based on cluster analysis. Instances from a validation set are firstly classified by all base classifiers. The classified results are then grouped into a number of clusters. Two instances are considered as being similar if they are correctly/incorrectly classified to the same class by the same group of classifiers. When classifying a new instance, the approach attempts to find the cluster to which it is closest. The method outperformed individual classifiers, classification by a clustering method and the majority voting method.
LanguageEnglish
Title of host publicationProceeding MCS'11 Proceedings of the 10th international conference on Multiple classifier systems
Pages259-268
Publication statusPublished - 2011

Fingerprint

Cluster analysis
Classifiers

Cite this

Jurek, A., Bi, Y., Wu, S., & Nugent, C. D. (2011). Classification by Cluster Analysis: A New Meta-Learning Based Approach. In Proceeding MCS'11 Proceedings of the 10th international conference on Multiple classifier systems (pp. 259-268)
Jurek, Anna ; Bi, Yaxin ; Wu, Shengli ; Nugent, Chris D. / Classification by Cluster Analysis: A New Meta-Learning Based Approach. Proceeding MCS'11 Proceedings of the 10th international conference on Multiple classifier systems. 2011. pp. 259-268
@inbook{13459bd27e284b82a8149b27931b5d08,
title = "Classification by Cluster Analysis: A New Meta-Learning Based Approach",
abstract = "Combination of multiple classifiers, commonly referred to as an classifier ensemble, has previously demonstrated the ability to improve classification accuracy in many application domains. One popular approach to building such a combination of classifiers is known as stacking and is based on a meta-learning approach. In this work we investigate a modified version of stacking based on cluster analysis. Instances from a validation set are firstly classified by all base classifiers. The classified results are then grouped into a number of clusters. Two instances are considered as being similar if they are correctly/incorrectly classified to the same class by the same group of classifiers. When classifying a new instance, the approach attempts to find the cluster to which it is closest. The method outperformed individual classifiers, classification by a clustering method and the majority voting method.",
author = "Anna Jurek and Yaxin Bi and Shengli Wu and Nugent, {Chris D.}",
year = "2011",
language = "English",
isbn = "978-3-642-21556-8",
pages = "259--268",
booktitle = "Proceeding MCS'11 Proceedings of the 10th international conference on Multiple classifier systems",

}

Jurek, A, Bi, Y, Wu, S & Nugent, CD 2011, Classification by Cluster Analysis: A New Meta-Learning Based Approach. in Proceeding MCS'11 Proceedings of the 10th international conference on Multiple classifier systems. pp. 259-268.

Classification by Cluster Analysis: A New Meta-Learning Based Approach. / Jurek, Anna; Bi, Yaxin; Wu, Shengli; Nugent, Chris D.

Proceeding MCS'11 Proceedings of the 10th international conference on Multiple classifier systems. 2011. p. 259-268.

Research output: Chapter in Book/Report/Conference proceedingChapter

TY - CHAP

T1 - Classification by Cluster Analysis: A New Meta-Learning Based Approach

AU - Jurek, Anna

AU - Bi, Yaxin

AU - Wu, Shengli

AU - Nugent, Chris D.

PY - 2011

Y1 - 2011

N2 - Combination of multiple classifiers, commonly referred to as an classifier ensemble, has previously demonstrated the ability to improve classification accuracy in many application domains. One popular approach to building such a combination of classifiers is known as stacking and is based on a meta-learning approach. In this work we investigate a modified version of stacking based on cluster analysis. Instances from a validation set are firstly classified by all base classifiers. The classified results are then grouped into a number of clusters. Two instances are considered as being similar if they are correctly/incorrectly classified to the same class by the same group of classifiers. When classifying a new instance, the approach attempts to find the cluster to which it is closest. The method outperformed individual classifiers, classification by a clustering method and the majority voting method.

AB - Combination of multiple classifiers, commonly referred to as an classifier ensemble, has previously demonstrated the ability to improve classification accuracy in many application domains. One popular approach to building such a combination of classifiers is known as stacking and is based on a meta-learning approach. In this work we investigate a modified version of stacking based on cluster analysis. Instances from a validation set are firstly classified by all base classifiers. The classified results are then grouped into a number of clusters. Two instances are considered as being similar if they are correctly/incorrectly classified to the same class by the same group of classifiers. When classifying a new instance, the approach attempts to find the cluster to which it is closest. The method outperformed individual classifiers, classification by a clustering method and the majority voting method.

M3 - Chapter

SN - 978-3-642-21556-8

SP - 259

EP - 268

BT - Proceeding MCS'11 Proceedings of the 10th international conference on Multiple classifier systems

ER -

Jurek A, Bi Y, Wu S, Nugent CD. Classification by Cluster Analysis: A New Meta-Learning Based Approach. In Proceeding MCS'11 Proceedings of the 10th international conference on Multiple classifier systems. 2011. p. 259-268