Eclectic method for feature reduction using Self-Organizing Maps

Author

DeLooze, Lori L.

Author_Institution

Comput. Sci. Dept., Naval Acad., Annapolis, MD

fYear

2008

fDate

1-8 June 2008

Firstpage

2069

Lastpage

2073

Abstract

This paper presents an eclectic method for extracting simple classification rules using a combination of a genetic algorithm, a self-organizing map and the ID3 decision tree algorithm. After outlining the method for extracting rules, we assess them for effectiveness, complexity and precision and compare them with similar methods which use support vector machines. While it is no surprise that the method proposed reduced the complexity of classification, it was surprising that the simple rules extracted from the SOMs were both more effective and more precise than the SOM from which they were extracted.

Keywords

genetic algorithms; self-organising feature maps; support vector machines; ID3 decision tree algorithm; classification rules extraction; eclectic method; feature reduction; genetic algorithm; self-organizing maps; support vector machines; Classification tree analysis; Decision trees; Feature extraction; Genetic algorithms; Genetic mutations; Pattern classification; Self organizing feature maps; Support vector machine classification; Support vector machines; Testing;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks, 2008. IJCNN 2008. (IEEE World Congress on Computational Intelligence). IEEE International Joint Conference on

Conference_Location

Hong Kong

ISSN

1098-7576

Print_ISBN

978-1-4244-1820-6

Electronic_ISBN

1098-7576

Type

conf

DOI

10.1109/IJCNN.2008.4634082

Filename

4634082