Analysis of diversity methods for evolutionary multi-objective ensemble classifiers

Stefan Oehmcke; Justin Heinermann; Oliver Kramer

doi:10.1007/978-3-319-16549-3_46

Analysis of diversity methods for evolutionary multi-objective ensemble classifiers

Stefan Oehmcke^*, Justin Heinermann, Oliver Kramer

^*Corresponding author for this work

2 Citations (Scopus)

Abstract

Ensemble classifiers are strong and robust methods for classification and regression tasks. Considering the balance between runtime and classifier accuracy the learning problem becomes a multi-objective optimization problem. In this work, we propose an evolutionary multiobjective algorithm based on non-dominated sorting that balances runtime and accuracy properties of nearest neighbor classifier ensembles and decision tree ensembles. We identify relevant ensemble parameters with a significant impact on the accuracy and runtime. In the experimental part of this paper, we analyze the behavior on typical classification benchmark problems.

Original language	English
Title of host publication	Applications of Evolutionary Computation - 18th European Conference, EvoApplications 2015, Proceedings
Editors	Giovanni Squillero, Antonio M. Mora
Number of pages	12
Publisher	Springer Verlag,
Publication date	1 Jan 2015
Pages	567-578
ISBN (Electronic)	9783319165486
DOIs	https://doi.org/10.1007/978-3-319-16549-3_46
Publication status	Published - 1 Jan 2015
Event	18th European Conference on the Applications of Evolutionary Computation, EvoApplications 2015 - Copenhagen, Denmark Duration: 8 Apr 2015 → 10 Apr 2015

Conference

Conference	18th European Conference on the Applications of Evolutionary Computation, EvoApplications 2015
Country/Territory	Denmark
City	Copenhagen
Period	08/04/2015 → 10/04/2015
Sponsor	Institute for Informatics and Digital Innovation, National Museum of Denmark, The World Federation on Soft Computing

Series	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	9028
ISSN	0302-9743

Keywords

Ensemble classification
Multi-objective optimization

Access to Document

10.1007/978-3-319-16549-3_46

Cite this

Oehmcke, S., Heinermann, J., & Kramer, O. (2015). Analysis of diversity methods for evolutionary multi-objective ensemble classifiers. In G. Squillero, & A. M. Mora (Eds.), Applications of Evolutionary Computation - 18th European Conference, EvoApplications 2015, Proceedings (pp. 567-578). Springer Verlag,. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) Vol. 9028 https://doi.org/10.1007/978-3-319-16549-3_46

Analysis of diversity methods for evolutionary multi-objective ensemble classifiers. / Oehmcke, Stefan; Heinermann, Justin; Kramer, Oliver.

Applications of Evolutionary Computation - 18th European Conference, EvoApplications 2015, Proceedings. ed. / Giovanni Squillero; Antonio M. Mora. Springer Verlag, 2015. p. 567-578 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 9028).

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

Oehmcke, S, Heinermann, J & Kramer, O 2015, Analysis of diversity methods for evolutionary multi-objective ensemble classifiers. in G Squillero & AM Mora (eds), Applications of Evolutionary Computation - 18th European Conference, EvoApplications 2015, Proceedings. Springer Verlag, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9028, pp. 567-578, 18th European Conference on the Applications of Evolutionary Computation, EvoApplications 2015, Copenhagen, Denmark, 08/04/2015. https://doi.org/10.1007/978-3-319-16549-3_46

Oehmcke S, Heinermann J, Kramer O. Analysis of diversity methods for evolutionary multi-objective ensemble classifiers. In Squillero G, Mora AM, editors, Applications of Evolutionary Computation - 18th European Conference, EvoApplications 2015, Proceedings. Springer Verlag,. 2015. p. 567-578. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 9028). doi: 10.1007/978-3-319-16549-3_46

Oehmcke, Stefan ; Heinermann, Justin ; Kramer, Oliver. / Analysis of diversity methods for evolutionary multi-objective ensemble classifiers. Applications of Evolutionary Computation - 18th European Conference, EvoApplications 2015, Proceedings. editor / Giovanni Squillero ; Antonio M. Mora. Springer Verlag, 2015. pp. 567-578 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 9028).

@inproceedings{0ac9d17da2a94d54be103493cfe99f06,

title = "Analysis of diversity methods for evolutionary multi-objective ensemble classifiers",

abstract = "Ensemble classifiers are strong and robust methods for classification and regression tasks. Considering the balance between runtime and classifier accuracy the learning problem becomes a multi-objective optimization problem. In this work, we propose an evolutionary multiobjective algorithm based on non-dominated sorting that balances runtime and accuracy properties of nearest neighbor classifier ensembles and decision tree ensembles. We identify relevant ensemble parameters with a significant impact on the accuracy and runtime. In the experimental part of this paper, we analyze the behavior on typical classification benchmark problems.",

keywords = "Ensemble classification, Multi-objective optimization",

author = "Stefan Oehmcke and Justin Heinermann and Oliver Kramer",

year = "2015",

month = jan,

day = "1",

doi = "10.1007/978-3-319-16549-3_46",

language = "English",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag,",

pages = "567--578",

editor = "Giovanni Squillero and Mora, {Antonio M.}",

booktitle = "Applications of Evolutionary Computation - 18th European Conference, EvoApplications 2015, Proceedings",

note = "18th European Conference on the Applications of Evolutionary Computation, EvoApplications 2015 ; Conference date: 08-04-2015 Through 10-04-2015",

}

TY - GEN

T1 - Analysis of diversity methods for evolutionary multi-objective ensemble classifiers

AU - Oehmcke, Stefan

AU - Heinermann, Justin

AU - Kramer, Oliver

PY - 2015/1/1

Y1 - 2015/1/1

N2 - Ensemble classifiers are strong and robust methods for classification and regression tasks. Considering the balance between runtime and classifier accuracy the learning problem becomes a multi-objective optimization problem. In this work, we propose an evolutionary multiobjective algorithm based on non-dominated sorting that balances runtime and accuracy properties of nearest neighbor classifier ensembles and decision tree ensembles. We identify relevant ensemble parameters with a significant impact on the accuracy and runtime. In the experimental part of this paper, we analyze the behavior on typical classification benchmark problems.

AB - Ensemble classifiers are strong and robust methods for classification and regression tasks. Considering the balance between runtime and classifier accuracy the learning problem becomes a multi-objective optimization problem. In this work, we propose an evolutionary multiobjective algorithm based on non-dominated sorting that balances runtime and accuracy properties of nearest neighbor classifier ensembles and decision tree ensembles. We identify relevant ensemble parameters with a significant impact on the accuracy and runtime. In the experimental part of this paper, we analyze the behavior on typical classification benchmark problems.

KW - Ensemble classification

KW - Multi-objective optimization

UR - http://www.scopus.com/inward/record.url?scp=84925873898&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-16549-3_46

DO - 10.1007/978-3-319-16549-3_46

M3 - Article in proceedings

AN - SCOPUS:84925873898

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 567

EP - 578

BT - Applications of Evolutionary Computation - 18th European Conference, EvoApplications 2015, Proceedings

A2 - Squillero, Giovanni

A2 - Mora, Antonio M.

PB - Springer Verlag,

T2 - 18th European Conference on the Applications of Evolutionary Computation, EvoApplications 2015

Y2 - 8 April 2015 through 10 April 2015

ER -

Analysis of diversity methods for evolutionary multi-objective ensemble classifiers

Abstract

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this