The comparison of optimizing SVM by GA and grid search

As we known, SVM is fit for the application of fault diagnosis. In our paper, we discussed the optimization methods for SVM. Including GA, Grid Search, and K-fold Cross Validation. For optimizing SVM, it is necessary to find out the best kernel function, to pick out the best kernel parameters and pe...

Full description

Saved in:
Bibliographic Details
Published in2017 13th IEEE International Conference on Electronic Measurement & Instruments (ICEMI) pp. 354 - 360
Main Authors Sun Yuanyuan, Wang Yongming, Guo Lili, Ma Zhongsong, Jin Shan
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.10.2017
Subjects
Computational complexity
Conferences
Genetic algorithms
Grid Search
K-fold cross validation
Kernel
Search methods
Support vector machines
SVM
Training
Online AccessGet full text

Cover

More Information
Summary:As we known, SVM is fit for the application of fault diagnosis. In our paper, we discussed the optimization methods for SVM. Including GA, Grid Search, and K-fold Cross Validation. For optimizing SVM, it is necessary to find out the best kernel function, to pick out the best kernel parameters and penalty factor parameters. Here, the standard datasets of UCI is used to illustrate the optimization effect by GA and Grid Search. The experiment environment is Matlab 2014a, the library of libsvm is adopted. The results can be seen that the Grid Search method has better performance than genetic algorithm in this circumstance.
DOI:10.1109/ICEMI.2017.8265815