| 自动文本分类中权值公式的改进 |
| |
| 引用本文： | 寇莎莎,魏振军.自动文本分类中权值公式的改进[J].计算机工程与设计,2005,26(6):1616-1618. |
| |
| 作者姓名： | 寇莎莎 魏振军 |
| |
| 作者单位： | 解放军信息工程大学,信息研究系,河南,郑州,450002 |
| |
| 摘 要： | 在自动文本分类中，TF-IDF公式是常用的词语权重计算公式，但是TF-IDF公式是一种经验公式，并没有坚实的理论基础，它并不适用于任何情况下。通过信息论和概率证明了，在训练文本同属一个类别时，词语的重要性与词语的文档频率成正比，并对TF-IDF进行了改进，得到了改进的权值公式。改进的权值公式与TF-IDF公式进行实验比较，实验结果表明改进的权值公式提高了算法的分类精度。
|
| 关 键 词： | 文本分类 TF-IDF 向量空间 特征项 特征权重 |
| 文章编号： | 1000-7024(2005)06-1616-03 |
Improved weighting formula in auto text classification |
| |
| KOU Sha-sha,WEI Zhen-jun.Improved weighting formula in auto text classification[J].Computer Engineering and Design,2005,26(6):1616-1618. |
| |
| Authors: | KOU Sha-sha WEI Zhen-jun |
| |
| Abstract: | In auto textclassification, TF-IDF is often used when the weight of a term is calculated. But it is only an experiential formula, so it is not employed at any time. By using information theory and probability is it proven, the importance of a term is in proportion to document frequency of the term. TF-IDF is improved, and an improved weighting formula is obtained. And better performance is achieved when texts of the training set are belonged to a category in the experiment and applications in a categorization system. |
| |
| Keywords: | text classification TF-IDF vector space model feature term term weighting |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
