矩阵奇异值分解理论在中文文本分类中的应用  

On the Application of Matrix Singular Value Decomposition Theory in Chinese Text Classification

作　　者：田东风[1] 欧飞[1] 申维[1] 

机构地区：[1]中国地质大学信息工程学院.北京100083

出　　处：《数学的实践与认识》2008年第24期132-140,共9页Mathematics in Practice and Theory

基　　金：国家自然科学基金(40172099;40672196)

摘　　要：用改进的截断与转换的矩阵奇异值分解算法,设计实现了基于字频特征的中文文本分类器.理论分析与实验结果表明,采用的方法提高了数值计算精度,降低了文本集特征空间的维数,简化了文本分类算法的时间复杂度,提高了文本分类准确率.By virtue of the improved truncation-transformation algorithm for the matrix singular value decomposition, we establish a Chinese text classification system based on the character frequency feature. The theoretical analysis and the experimental results indicate that the method used in this paper for designing the Chinese text classification can improve the precise of the numerical computation, decrease the dimension of the characteristic space of the text set, simplify the time complexity of the existed algorithm for the text classification, ameliorate the accuracy of the text classification and the like.

关 键 词：奇异值分解 文本分类 特征选择 

分 类 号：O151.21[理学—数学]