Using a neural network and genetic algorithm to extract decision rules

Karen Blackmore; Terence Bossomaier

Using a neural network and genetic algorithm to extract decision rules

Research output: Book chapter/Published conference paper › Conference paper › peer-review

40 Downloads (Pure)

Abstract

Rule extraction from neural networks often focusses on exact equivalence and is often tested on relatively small canonical examples. We apply genetic algorithms to the extract approximate rules from neural networks. The method is robust and works with large networks. We compare the results with rules obtained using state of the art decision tree methods and achieve superior performance to straight forward application of the WEKA implementation of the C5 algorithm, J48.PART

Original language	English
Title of host publication	Combined proceedings of DICTA 2003 and ANZIIS 2003
Editors	Brian C Lovell, Duncan A Campbell, Clinton B Fookes, Anthony J Maeder
Place of Publication	Brisbane
Publisher	Queensland University of Technology
Pages	187-191
Number of pages	5
ISBN (Electronic)	1741070392
Publication status	Published - 2003
Event	Australian and New Zealand Intelligent Information Systems Conference - Sydney, Australia, Australia Duration: 10 Dec 2003 → 12 Dec 2003

Conference

Conference	Australian and New Zealand Intelligent Information Systems Conference
Country/Territory	Australia
Period	10/12/03 → 12/12/03

Access to Document

Using a neural network and genetic algorithm to extract decision rulesAccepted author manuscript, 81.1 KB

Cite this

@inproceedings{ade5f717986a453fafe538a60e3de979,

title = "Using a neural network and genetic algorithm to extract decision rules",

abstract = "Rule extraction from neural networks often focusses on exact equivalence and is often tested on relatively small canonical examples. We apply genetic algorithms to the extract approximate rules from neural networks. The method is robust and works with large networks. We compare the results with rules obtained using state of the art decision tree methods and achieve superior performance to straight forward application of the WEKA implementation of the C5 algorithm, J48.PART",

keywords = "Open access version available",

author = "Karen Blackmore and Terence Bossomaier",

note = "Imported on 03 May 2017 - DigiTool details were: publisher = Brisbane: Queensland University of Technology, 2003. editor/s (773b) = Lovell, Brian C ; Campbell, Duncan A ; Fookes, Clinton B ; Maeder, Anthony J ; Event dates (773o) = 10-12 December 2003; Parent title (773t) = Australian and New Zealand Intelligent Information Systems Conference.; Australian and New Zealand Intelligent Information Systems Conference ; Conference date: 10-12-2003 Through 12-12-2003",

year = "2003",

language = "English",

pages = "187--191",

editor = "Lovell, {Brian C} and Campbell, {Duncan A} and Fookes, {Clinton B} and Maeder, {Anthony J}",

booktitle = "Combined proceedings of DICTA 2003 and ANZIIS 2003",

publisher = "Queensland University of Technology",

address = "Australia",

}

Using a neural network and genetic algorithm to extract decision rules. / Blackmore, Karen; Bossomaier, Terence.
Combined proceedings of DICTA 2003 and ANZIIS 2003. ed. / Brian C Lovell; Duncan A Campbell; Clinton B Fookes; Anthony J Maeder. Brisbane: Queensland University of Technology, 2003. p. 187-191.

Research output: Book chapter/Published conference paper › Conference paper › peer-review

TY - GEN

T1 - Using a neural network and genetic algorithm to extract decision rules

AU - Blackmore, Karen

AU - Bossomaier, Terence

N1 - Imported on 03 May 2017 - DigiTool details were: publisher = Brisbane: Queensland University of Technology, 2003. editor/s (773b) = Lovell, Brian C ; Campbell, Duncan A ; Fookes, Clinton B ; Maeder, Anthony J ; Event dates (773o) = 10-12 December 2003; Parent title (773t) = Australian and New Zealand Intelligent Information Systems Conference.

PY - 2003

Y1 - 2003

N2 - Rule extraction from neural networks often focusses on exact equivalence and is often tested on relatively small canonical examples. We apply genetic algorithms to the extract approximate rules from neural networks. The method is robust and works with large networks. We compare the results with rules obtained using state of the art decision tree methods and achieve superior performance to straight forward application of the WEKA implementation of the C5 algorithm, J48.PART

AB - Rule extraction from neural networks often focusses on exact equivalence and is often tested on relatively small canonical examples. We apply genetic algorithms to the extract approximate rules from neural networks. The method is robust and works with large networks. We compare the results with rules obtained using state of the art decision tree methods and achieve superior performance to straight forward application of the WEKA implementation of the C5 algorithm, J48.PART

KW - Open access version available

M3 - Conference paper

SP - 187

EP - 191

BT - Combined proceedings of DICTA 2003 and ANZIIS 2003

A2 - Lovell, Brian C

A2 - Campbell, Duncan A

A2 - Fookes, Clinton B

A2 - Maeder, Anthony J

PB - Queensland University of Technology

CY - Brisbane

T2 - Australian and New Zealand Intelligent Information Systems Conference

Y2 - 10 December 2003 through 12 December 2003

ER -

Using a neural network and genetic algorithm to extract decision rules

Abstract

Conference

Access to Document

Fingerprint

Cite this