基于类别方差的特征权重算法

特征权重计算是文本分类过程的基础，传统基于概率的特征权重算法，往往只对词频，逆文档频和逆类频等进行统计，忽略了类别之间的相互关系。而对于多分类问题，类别之间的关系对统计又有重要意义。因此，针对这一不足，本文提出了基于类别方差的特征权重算法，通过计算类别文档频率的方差来度量类别之间的联系，并在搜狗新闻数据集上对五种特征权重算法进行分类实验。结果表明，与其他四种特征权重算法相比，本文提出的算法在F1宏平均和F1微平均上都有较大的提高，提升了文本分类的效果。

A feature weighting algorithm based on class variance

The computation of feature weight is the basis of text categorization, and the traditional feature weighting algorithm based on probability always counts only word frequency, inverse document frequency and inverse class frequency, ignoring the relation among classes. For multi classification problems, the relationship between classes is important to statistics. Therefore, to solve this problem, this paper proposed a feature weight algorithm based on class variance to measure the relationship between categories, through variance category document frequency, and classification experiments on five kinds of feature weight algorithm in the data set on the Sogou news. The results show that compared with the other four feature weighting algorithms, the proposed algorithm has greatly improved the F1 macro mean and the F1 micro mean, and improved the effect of text categorization.