Proceedings of the 2007 International Conference on Intelligent Systems and Knowledge Engineering (ISKE 2007)

Evaluation of the quality of attribute reductions obtained using rough set by margin criteria

Authors
Ming Yang1, Xuqiang Qiu, Genlin Ji
1Department of Computer Science, Nanjing Normal University
Corresponding Author
Ming Yang
Available Online October 2007.
DOI
10.2991/iske.2007.116How to use a DOI?
Keywords
Rough Set; Attribute Reduction; Margin
Abstract

As a filter model, rough set-based methods are one of effective attribute reduction(also called feature selection) that preserve the meaning of the features. In rough set theory, researchers mainly focus on extension of the classical rough set model(also called Pawlak model for short) and development of efficient attribute reduction algorithms. However, very little work has been done for aiming on the evaluation of the quality of attribute reduction, except that employing the cardinality of the given attribute subset P and so-called approximation quality of P or other equivalent criteria induced by Pawlak model. Although this discrimination strategy is simple and effective in most cases, it is very difficulty to guarantee the selected attribute reduction(s) from lots of attribute reductions are the best or Top n, especially for the case containing many attribute reductions with same cardinality and approximation quality. Therefore, in this paper, we incorporate margin criteria into the proposed evaluation mechanism for guaranteeing the effectiveness of the selected attribute subsets, since margin, originally designed for binary classification problem using support vector machine, can actually determine the generalization ability. Also, an improved discernibility function-based algorithm is proposed. To further test the effectiveness of the proposed method, the algorithm of this paper is experimented using UCI benchmark datasets. Preliminary experimental results show that the attribute reductions with larger margin have better or comparable performance than those with relatively small margin for all reducts with same cardinality. Thus, our newly developed method can, in most cases, get more effective attribute subsets.

Copyright
© 2007, the Authors. Published by Atlantis Press.
Open Access
This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

Volume Title
Proceedings of the 2007 International Conference on Intelligent Systems and Knowledge Engineering (ISKE 2007)
Series
Advances in Intelligent Systems Research
Publication Date
October 2007
ISBN
978-90-78677-04-8
ISSN
1951-6851
DOI
10.2991/iske.2007.116How to use a DOI?
Copyright
© 2007, the Authors. Published by Atlantis Press.
Open Access
This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Cite this article

TY  - CONF
AU  - Ming Yang
AU  - Xuqiang Qiu
AU  - Genlin Ji
PY  - 2007/10
DA  - 2007/10
TI  - Evaluation of the quality of attribute reductions obtained using rough set by margin criteria
BT  - Proceedings of the 2007 International Conference on Intelligent Systems and Knowledge Engineering (ISKE 2007)
PB  - Atlantis Press
SP  - 676
EP  - 682
SN  - 1951-6851
UR  - https://doi.org/10.2991/iske.2007.116
DO  - 10.2991/iske.2007.116
ID  - Yang2007/10
ER  -