Proceedings of the 2016 4th International Conference on Mechanical Materials and Manufacturing Engineering

Research on Sample Dataset Balance Method of SVM Based on GA

Authors
Xiao Han
Corresponding Author
Xiao Han
Available Online October 2016.
DOI
10.2991/mmme-16.2016.197How to use a DOI?
Keywords
SVM, Fault Diagnosis, Sample Balance, GA
Abstract

SVM was widely used in fault diagnosis, and achieved good results. However, the unbalance between normal sample datasets and fault sample datasets made it very difficult to establish a proper diagnosis model. For ac-tual diagnosis, the normal samples are usually more than the fault ones, and it will lead to misdiagnosis. In this paper, a method based on GA to solve the imbalance problem for SVM is presented. In this method, the sam-ples are expanded by GA so that the number of normal sample datasets and fault sample datasets keeps bal-ance. The method of selecting parent samples is also studied. The experiments show that the method proposed in this paper improves the accuracy of diagnosis.

Copyright
© 2016, 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 2016 4th International Conference on Mechanical Materials and Manufacturing Engineering
Series
Advances in Engineering Research
Publication Date
October 2016
ISBN
10.2991/mmme-16.2016.197
ISSN
2352-5401
DOI
10.2991/mmme-16.2016.197How to use a DOI?
Copyright
© 2016, 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  - Xiao Han
PY  - 2016/10
DA  - 2016/10
TI  - Research on Sample Dataset Balance Method of SVM Based on GA
BT  - Proceedings of the 2016 4th International Conference on Mechanical Materials and Manufacturing Engineering
PB  - Atlantis Press
SN  - 2352-5401
UR  - https://doi.org/10.2991/mmme-16.2016.197
DO  - 10.2991/mmme-16.2016.197
ID  - Han2016/10
ER  -