SVR Kernel Parameters Selection Based on Steady-State Genetic Algorithm

Author

Li, Jie ; Gao, Feng ; Guan, Xiaohong ; Xu, Hui

Author_Institution

Syst. Eng. Inst., Xi´´an Jiaotong Univ.

Volume

1

fYear

0

fDate

0-0 0

Firstpage

4405

Lastpage

4409

Abstract

The hyper parameters selection has a great affection on the accuracy of support vector regression algorithm. We chose the optimal hyper parameters including kernel parameters based on steady genetic algorithm for the support vector regression model. Selection of usually used RBF kernel parameters was thoroughly investigated. Two selection strategies for single and diagonal kernel parameters selection were applied on the standard sample data for Boston housing forecasting, and for electrical power demand forecasting. The testing results show that applying steady GA is effective in selecting multiple parameters

Keywords

genetic algorithms; parameter estimation; regression analysis; support vector machines; RBF kernel parameter; SVR kernel parameter selection; diagonal kernel parameter selection; hyper parameter selection; optimal hyper parameter; radial basis function; single kernel parameter selection; steady-state genetic algorithm; support vector regression; Electronic mail; Genetic algorithms; Genetic engineering; Intelligent networks; Intelligent systems; Kernel; Load forecasting; Power system modeling; Steady-state; Systems engineering and theory; Hyper Parameters selection; Steady-state Genetic Algorithm; Support Vector Machine; kernel Parameters;

fLanguage

English

Publisher

ieee

Conference_Titel

Intelligent Control and Automation, 2006. WCICA 2006. The Sixth World Congress on

Conference_Location

Dalian

Print_ISBN

1-4244-0332-4

Type

conf

DOI

10.1109/WCICA.2006.1713210

Filename

1713210