基于类约束的贝叶斯网络分类器学习_王双成.pdfVIP

25 6 Vol.25 No.6 2004 6 MINI- MICRO SYST EMS June 2004 1 1 2 王双成, 苑森淼, 王　辉 1( , 130025) 2( , 130024) : 分类能力是人类经过学习得到的重要而基本的能力, 也是机器学习、模式识别和数据采掘 究的核心问题.在 0-1 损失率下, 证明了基于类约束的贝叶斯网络分类器是最优分类器.建立该分类器的核心问题是基于类约束属性贝 叶斯网络结构学习, 给出了学习属性贝叶斯网络结构的方法, 在学习过程中使用了根据弧方向因果语义确定边方向的 方法, 并和碰撞识别定向相结合,在边定向之后进行冗余弧检验, 解决了目前冗余边检验在定向之前所导致的问题,显 著提高了结构学习效率和准确性. 并使用模拟数据进行了分类实验和分析. : ;0-1 ; ; :T 391:A : 1000-1220( 2004) -06-0968-04 Learning Bayesian Network Classif iers Restricted by Class Variable - 1, - 1, 2 WANG Shuang cheng YUAN Sen miao WANG Hui 1( Institute of Comp uter Science and Technology, Jilin University, Changchun 130025, China) 2( Dep artment of Comp uter Science N ortheast University Changchun, 130024, China) Abstract: The classificat ion is an important and basic ability for human obt ained by learning It has been considered as. a key research area in machine learning, pattern recognit ion and dat a mining. It is proved that a Bayesian network clas- - . sifier restrict ed by class variable is optimal under zero one loss rate T he most im portant problem of setting up the classifier is to learning the structure of attributes Bayesian network restricted by class variable. In this paper, the method of learning the structure of attributes Bayesian netw ork is developed In learning t he method of orienting edges. based on the causal semanitics of an arcs direction is used. The method is combined w ith that of orienting edges based on collider identificat ion to make superfluous arcs disposal after orienting edges