基于词条聚合和决策树的文本分类方法  

Text Categorization Based on Word Aggregation and Decision Tree

作　　者：王煜[1] 张明[1] 马力[2] 

机构地区：[1]河北大学数学与计算机学院,河北保定071002 [2]河北大学出版社,河北保定071002

出　　处：《河北大学学报（自然科学版）》2005年第3期338-342,共5页Journal of Hebei University(Natural Science Edition)

基　　金：河北省科学技术研究与发展计划项目(04213534)

摘　　要：根据词条聚合和决策树原理,提出了一种文本分类的新方法.决策树分类方法具有出色的数据分析效率和容易抽取易于理解的分类规则等优势,但只能应用于维数较低的特征空间.本方法将与各个类别相关程度相似的词条聚合为一个特征,有效地降低了向量空间的维数,然后再使用决策树进行分类,从而既保证了分类精度又获得了决策树易于抽取分类规则的优势.A new method of text classifying by using the theory of word aggregation and decision tree is presenteal. The decision tree is applied to text categorization, which has the advantages of high efficiency of data analysis and easily abstracting the categorization rules that are able to understand. However, decision tree has a defect that is only suitable for low dimension of features. The new method establishes the vector space model of term weight in terms of the theory of Pattern Aggregation, which merges such words as a new feature that has the similar mutuality with each class, and so largely reduces the dimension of the vector space. After that, the decision tree is applied to text categorization. Both the advantage of decision tree and better accuracy of categorization can be acquired.

关 键 词：文本分类 互信息 决策树 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]