一种改进的支持向量机的文本分类算法_巩知乐.pdfVIP

26 7 2009 7 : 1006 - 9348( 2009) 07 - 0165- 04 1 1 2 巩知乐 , 张德贤 , 胡明明 ( 1. , 450001 2. , 710071) : , ( SVM ) , ( IA - SVM ) , SVM , IA- SVM , : ; ; : TP311; TP301 : A An Improved SVM Algorithm forChineseTextClassification 1 1 2 GONG Zh i- l , ZHANG D - x ian , HU M ing- m ing ( 1. School of In form ation Sci nc and Engin ring, H nan Un iv rsity of T chnology, Zh ngzhou H nan 45000 1, Ch ina; 2. School of Com put r, X i. an Univ rs ity of E l ctronic Sci nc and T chnology, X i. an Shanx i 710071, China) ABSTRACT: In th t xt class ification fi ld, using SupportV ctorM ach in s ( SVM ) algorithm can ob tain a satisfacto- ry g n ralization ab ility of classification und r th condition of small samp l s. . But th param t rs of th Support V ctor M ach in s d cid its l arn ing p rform anc and g n ralization ab ility. T o nhanc th p rformanc of Support V ctor M ach in s ( SVM ) algorithm, th is mi prov d t xt classification algorithm ( IA - SVM ) us s mi m unity algorithm ( IA) to optmi iz th param t rs of trad itional SVM algorithm un ion, thus d cr as ing th blindn ss in d t rm ining SVM param t rs, and nhancing th pr cision of pr d iction of SVM . Th xp rmi nt indicat s that this mi prov d IA - SVM algorithm can rais th accuracy of class ification significantly, and ob tains a satisfactory acc l ration in th l arning proc ss. KEY ORDS: Immun algorithm ( IA); Support v ctorm achin s( SVM ); T xt classification( TC) 1 , In t rn t, [ 2] [ 3] ,