关键词: 文本分类 层次式分类 Naive Bayes 机器学习 数据偏斜

Abstract:

The text categorization method of Naive Bayes, which highly depends on the subjectively-selected sample distribution with respect to classes, is enhan ced by using the characteristics of hierarchical classification and introducing the conditional probability. The enhancement makes decisions in the loca l data which belong to the sub-classes of an internal class to lighten the influence of global data distribution, and partially overcomes the problem ofdata skewness. Experiments show the enhanced method improves the effectiveness of hierarchical categorization with the Naive Bayes method notably.

Key words: text categorization, hierarchical categorization, Naive Bayes, machine learning, data skewness