广义朴素贝叶斯分类器  

GENERALIZED NAIVE BAYES CLASSIFIER

作　　者：王双成[1] 忻瑞婵[1] 

机构地区：[1]上海立信会计学院信息科学系

出　　处：《计算机应用与软件》2007年第11期12-13,20,共3页Computer Applications and Software

基　　金：国家自然科学基金(60275026);上海市重点学科(P1601);上海市教委重点项目(05zz66)资助

摘　　要：朴素贝叶斯分类器具有很高的学习和分类效率,但不能充分利用属性变量之间的依赖信息。贝叶斯网络分类器具有很强的分类能力,但分类器学习比较复杂。本文建立广义朴素贝叶斯分类器,它具有灵活的分类能力选择方式、效率选择方式及学习方式,能够弥补朴素贝叶斯分类器和贝叶斯网络分类器的不足,并继承它们的优点。Naive Bayes classifier has high learning and classing efficiency, but it can not make the best use of the dependent information between attribute variates （the dependent information is also important classing information）. Bayesian network classifier has high classing a- bility, but it is very difficult to learn. The generalized naive Bayes classifier is presented, which has flexible manner of choosing classing ability, efficiency and learning classifier from data. It makes up the deficiency of Naive Bayes and Bayesian network classifiers and inherits their advantages.

关 键 词：广义朴素贝叶斯分类器 贪婪搜索 分类能力 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]