CAJ | 학술논문

多项式核函数由于具有良好的泛化性能而受到重视,并被研究用于文本分类问题.针对多项式核学习能力较差的缺点,将学习能力较强的条件正定核与多项式核构成一个混合核函数作为改进的多项式核.实验表明,改进的多项式核SVM文本分类器的分类效果要好于多项式核SVM文本分类器.
다항식핵함수유우구유량호적범화성능이수도중시,병피연구용우문본분류문제.침대다항식핵학습능력교차적결점,장학습능력교강적조건정정핵여다항식핵구성일개혼합핵함수작위개진적다항식핵.실험표명,개진적다항식핵SVM문본분류기적분류효과요호우다항식핵SVM문본분류기.
Polynomial kernel is investigated and widely used for text categorization because of it' s high generalization performance. For the polynomial kernel low study performance' s fault, this paper combined conditionally positive definite kernel which had high study performance with polynomial kernel as an improved polynomial kernel. The experiment results show that the improved polynomial kernel SVM classifier for text categorization is superior to polynomial kernel SVM classifier for text categorization.