基于属性相关性分析的贝叶斯分类模型  

Bayesian Classification Model Based on Attribute Correlation Analysis

作　　者：章舜仲[1] 王树梅[1] 黄河燕[2] 陈肇雄[2] 

机构地区：[1]南京理工大学计算机科学系,南京210094 [2]中国科学院计算机语言信息工程研究中心,北京100083

出　　处：《情报学报》2007年第2期271-274,共4页Journal of the China Society for Scientific and Technical Information

摘　　要：朴素贝叶斯分类器是一种简单而有效的概率分类方法，然而其属性独立性假设在现实世界中多数不能成立。为改进其分类性能，近几年已有大量研究致力于构建能反映属性之间依赖关系的模型。本文提出一种向量相关性度量方法，特征向量属于类的的概率由向量相关度及其属性概率计算。向量相关度可通过本文给出的一个公式进行估计。实验结果表明，使用这种方法构建的分类模型其分类性能明显优于朴素贝叶斯，和其他同类算法相比也有一定提高。Naive Bayes classifier is a simple and effective classification method based on probability theory, but its attribute independence assumption is often violated in the real world. To improve the performance of Bayes classifiers, in recent years, a great deal of research has been done on constructing models which can express dependence among attributes. This paper presented a method for measaring the correlation of a vector. The probability of a character vector belonging to a class is calculated by vector＇s correlation degree and the probability of its properties, and the vector correlation degree can be computed via a formula given in the paper. Experiments showed that the classifier built by this method achieved higher accuracy than NB and other similar algorithm.

关 键 词：分类模型 贝叶斯定理 属性相关 向量相关度 

分 类 号：O212.1[理学—概率论与数理统计]