基于条件互信息下聚类的朴素贝叶斯分类算法

Naive Bayesian classification algorithm based on clustering with conditional mutual information

摘要: 采用条件互信息来度量任意2个条件属性之间的关联程度,采用互信息度量各条件属性与类属性间的关联程度,以此作为将各条件属性进行聚类的准则,提出一种新的将条件属性进行聚类的分组技术.同时,结合朴素贝叶斯分类算法,构造了改进的朴素贝叶斯分类模型.通过仿真实验表明该文提出的算法具有较好的分类性能.

Abstract: In this paper,the correlation intensity of two arbitrary conditional attributes was measured by conditional mutual information,and the correlation intensity between every conditional attribute and classification attribute was measured by mutual information.On that criterion to cluster the conditional attributes,a new grouping method to cluster the conditional attributes was proposed.Simultaneously,combined with naive bayes classification algorithm,an improved naive bayes classification model was constructed.Simulation results showed the efficiency of this method is preferable.