| 中文文本分类中的特征选择研究 |
| |
| 引用本文： | 寇苏玲,蔡庆生.中文文本分类中的特征选择研究[J].计算机仿真,2007,24(3):289-291. |
| |
| 作者姓名： | 寇苏玲 蔡庆生 |
| |
| 作者单位： | 中国科学技术大学计算机系,安徽,合肥230027 |
| |
| 摘 要： | 有多种特征选择算法被用于文本自动分类,YimingYang教授曾针对英文文本分类中的特征选择做过深入的研究,并得出结论:IG和CHI方法效果相对较好.考虑到该结论不一定适合对中文文本的分类,对中文文本分类中的特征选择方法进行研究,采用了包含500篇新闻的中文语料库对几种特征选择算法进行测试,结果表明:在测试的特征选择算法中,χ2估计方法无需因训练集的改变而人为调节特征阀值,并且分类准确率较高.
|
| 关 键 词： | 特征选择 特征提取 文本分类 中文文本分类 特征选择算法 选择研究 Chinese Text Classification 分类准确率 阀值 调节 训练集 估计方法 结果 测试 中文语料库 新闻 选择方法 效果 英文 自动 |
| 文章编号： | 1006-9348(2007)03-0289-03 |
| 修稿时间： | 2005-11-30 |
Research on Feature-Selection in Chinese Text Classification |
| |
| KOU Su-ling,CAI Qing-sheng.Research on Feature-Selection in Chinese Text Classification[J].Computer Simulation,2007,24(3):289-291. |
| |
| Authors: | KOU Su-ling CAI Qing-sheng |
| |
| Affiliation: | Department of Computer Science, USTC, Hefei Anhui 230027, China |
| |
| Abstract: | Several algorithms of feature-selection are used in text-classification.Professor YumingYang has deeply studied the feature-selection in English text-classification,and got a conclusion that the IG and CHI are better than the others.Because Chinese is more complicated than English,this paper is about a research into feature -selection in Chinese text-classification.The results of the experiments based on 500 news articles indicate that of all the algorithms the X~2 statistic(CHI)has the highest precision.And when the size of the training set is changed, manual threshold-adjustment is not required. |
| |
| Keywords: | Feature-selection Feature-abstraction Text-classification |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
