Computer Engineering and Applications ›› 2011, Vol. 47 ›› Issue (3): 38-40.DOI: 10.3778/j.issn.1002-8331.2011.03.011

• 研究、探讨 • Previous Articles     Next Articles

Method of determining Gaussian kernel parameter by clustering

LIU Qiongsun,FAN Ruiya   

  1. College of Mathematics and Physics,Chongqing University,Chongqing 400030,China
  • Received:2010-05-05 Revised:2010-07-23 Online:2011-01-21 Published:2011-01-21
  • Contact: LIU Qiongsun

确定高斯核参数的聚类方法

刘琼荪,范瑞雅   

  1. 重庆大学 数理学院,重庆 400030
  • 通讯作者: 刘琼荪

Abstract: The selection of Gaussian kernel parameterσdirectly affects the classification of Gaussian kernel SVM(Support Vector Machine).Togethering the clustering method and minimum distance,an optimization algorithm is constructed to determine the parameterσ,which adapts Gaussian kernel SVM to classify the test set,whose correct classification rate contributes the selection of parameterσ.Experimental results show that this method is suitable for a broader data type,furthermore,the method has a good generalization ability,and efficiently develops results of the classification.

Key words: Gaussian kernel parameter, clustering, minimum distance, Support Vector Machine(SVM)

摘要:

高斯核参数σ的选择,直接影响着高斯核支持向量机的分类性能。将聚类方法与最小距离分类法进行融合,构造了能有效确定高斯核参数σ的优化算法。采用高斯核支持向量机方法对测试集进行分类,以分类正确率来评判选取核参数σ的效果。实验表明,该方法适宜于较广泛的数据类型,具有良好的推广能力,并能有效提高分类效果。

关键词: 高斯核参数, 聚类, 最小距离, 支持向量机

CLC Number: