| 文本分类中特征选择方法的比较和改进 |
| |
| 引用本文： | 伍建军,康耀红.文本分类中特征选择方法的比较和改进[J].郑州大学学报(理学版),2007,39(2):110-113. |
| |
| 作者姓名： | 伍建军 康耀红 |
| |
| 作者单位： | 海南大学信息科学技术学院,海口,570228 |
| |
| 摘 要： | 考察了文档频率DF、互信息MI、CHI统计、CC统计四种不同的特征选择方法,并结合K近邻算法进行分类精度上的比较.为消除MI对低频词的倚重,提出一种DF与MI结合的特征评价函数,并验证了这种组合特征选择方法的有效性.
|
| 关 键 词： | 特征选择 互信息 文本分类 |
| 文章编号： | 1671-6841(2007)02-0110-04 |
| 修稿时间： | 12 20 2006 12:00AM |
Comparison and Improvement of Feature Selection for Text Categorization |
| |
| WU Jian-jun,KANG Yao-hong.Comparison and Improvement of Feature Selection for Text Categorization[J].Journal of Zhengzhou University:Natural Science Edition,2007,39(2):110-113. |
| |
| Authors: | WU Jian-jun KANG Yao-hong |
| |
| Abstract: | Four different feature selection methods are discussed,including Document Frequency(DF),Mutual Information(MI),X2 test(CHI),Correlation Coefficient(CC),and the correction of text categorization is compared using the algorithm of K nearest neighbor.In order to decrease the MI's bias towards low frequency words,one combined feature selection is expounded,and is proved to be effective by experiment. |
| |
| Keywords: | feature selection mutual information text categorization |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
