一种改进的不完备信息系统分类器  

An Improved Bayesian Classifier for Incomplete Information Systems

作　　者：杨景 宋丽 YANG Jing;SONG Li(Faculty of Electrical and Mechanical Engineering,Kunming Metallurgy College,Kunming 650033,China;China Mobile Communications Group Yunnan Co.,Ltd.,Yuxi Branch,Yuxi 653100,Yunnan,China)

机构地区：[1]昆明冶金高等专科学校电气与机械学院,云南昆明650033 [2]中国移动通信集团云南有限公司玉溪分公司,云南玉溪653100

出　　处：《昆明冶金高等专科学校学报》2022年第6期77-83,共7页Journal of Kunming Metallurgy College

摘　　要：分类是数据挖掘的一个重要功能,贝叶斯算法是众多分类算法中一个经典算法,具有极高的准确率。然而,贝叶斯算法要处理的数据必须具备完备的信息系统。为建立不完备信息系统的分类模型,改进了已有的针对不完备信息系统的DBCI分类器,构造了IBCI分类器。该分类器不需要事前对数据进行补齐操作,可避免因数据补齐导致的分类模型准确率较低问题。理论和试验均证明该分类器比改进前的DBCI分类器具有更高精度,拓展了贝叶斯算法的适用范围。Classification is an important data mining function. Featuring high accuracy,Bayesian algorithm is one of the classical classification algorithms. However, Bayesian algorithm requires that the data to be processed is a complete information system, and the algorithm cannot be directly applied to incomplete information systems. In order to establish a classification model for incomplete information systems, this paper improves the DBCI classifier and constructs the IBCI classifier, which does not need to supplement the data in advance. The classifier avoids the low accuracy of classification model caused by data complement. Both the theory and the experimental results prove that the classifier has higher accuracy than the DBCI classifier, and extends the application scope of Bayesian algorithm.

关 键 词：数据挖掘 分类 不完备信息系统 贝叶斯算法 

分 类 号：TN919.5[电子电信—通信与信息系统]