Abstract: The feature vector of Chinese Web page is high dimension and very sparse for text categorization.How to reduce the dimensionality of feature space is a very key problem for practical text classification.In this paper a new method is described.The approach is to take advantage of latent semantic analysis and feature selection that use statistical methods.The K-Nearest Neighbor method is selected as the evaluating classifiers.The experimental result shows that the proposed method for Chinese Web page categorization to be promising.

Key words: Web Page categorization, latent semantic analysis, feature selection, KNN

摘要：

基于向量空间模型的文本分类中特征向量是极度稀疏的高维向量，只有降低向量空间维数才能提高分类效率。在利用统计方法选择文本分类特征降低特征空间维数的基础上，采用隐含语义分析技术，挖掘文档特征间的语义信息，利用矩阵奇异值分解理论进一步降低了特征空间维数。实验结果表明分类结果宏平均F1约提高了5%，验证了该方法的有效性。

关键词: 网页分类, 隐含语义分析, 特征选择, KNN