基于Rough Set的加权朴素贝叶斯分类算法  

Weighted Naive Bayes Classification Algorithm Based on Rough Set

作　　者：邓维斌[1] 王国胤[1] 王燕[1] 

机构地区：[1]重庆邮电大学计算机科学与技术研究所

出　　处：《计算机科学》2007年第2期204-206,219,共4页Computer Science

基　　金：国家自然科学基金(No.60373111);新世纪优秀人才支持计划;重庆市教委科学技术研究项目(040505)资助

摘　　要：朴素贝叶斯算法是一种简单而高效的分类算法,但其条件独立性假设并不符合客观实际,这在某种程度上影响了它的分类性能。加权朴素贝叶斯是对它的一种扩展。基于Rough Set的属性重要性理论,提出了基于Rough Set的加权朴素贝叶斯分类方法,并分别从代数观、信息观及综合代数观和信息观的角度给出了属性权值的求解方法。通过在UCI数据集上的仿真实验,验证了该方法的有效性。Naive Bayes algorithm is an effective simple classification algorithm. Since its conditional independence as sumption is not always true in real life, its classification performance is affected to some extent. Weighted naive Bayes （simply WNB） is an extension of it. Based on the attributes＇ importance degree theory of rough set, a new weighted naive Bayes method is proposed. Methods for determining the weights of attributes in the algebra view, informational view and both of them are developed respectively. Simulation results on a variety of UCI data sets illustrate the efficien cy of this method.

关 键 词：朴素贝叶斯 加权朴素贝叶斯 ROUGH集 属性重要性 分类 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]