一种改进的基于神经网络的文本分类算法*

提出并实现了一种结合前馈型神经网络和K最近邻的文本分类算法。其中，在选取特征项时考虑到Web文本不同标签组所代表的意义和权重有所区别，采用了一种改进的TFIDF特征选择法。最后对设计的分类器进行了开放性测试，实验结果表明该分类器显著地提高了文本分类的查全率和查准率。

Improved text classification method based on neural network

This paper put forward and carried out a text classification method using feed-forward neural network and K-nearest neighbor.Considering the differences of meaning and weight among the diverse groups of label in Web document,this paper adopted an improved TFIDF feature selection method.Finally an open test was done on this classifier. The experiment results show that this classifier can remarkably upgrade the recall and precision of text classification.