基于属性相关性分析的扩展朴素贝叶斯分类器

朴素贝叶斯分类作为一种统计分类的方法,简单高效,但它是建立在属性独立性假设的基础上,有一定的局限性,影响了它的分类效果.x2统计是一种度量属性相关性的方法,通过属性相关的分析,可以对属性进行约简,去除冗余和无关属性,达到简化朴素贝叶斯分类器的目的.朴素贝叶斯分类器的扩展方法是在非类父子结点之间添加扩展弧,表示相关属性间的依赖关系,从而扩展朴素贝叶斯分类器的结构.在数据集上的实验结果显示,KEANBC分类器优于NBC分类器,提高了分类正确率.

Extended Naive Bayesian Classifier Based on Attribute Correlation Analysis

Naive Bayesian classifier is a statistical classification method. The method is simple and efficient,but it is based on the assumption of attribute independence and has certain limitations,which affects its classification effect. x2 Statistics is a method of measuring the correlation of attributes. Through attribute correlation analysis,attributes can be reduced,redundant and unrelated attributes can be removed,and the purpose of simplifying the Naive Bayesian classifier can be achieved. The extension method of the Naive Bayesian classifier is to add an extension arc between non-class parent nodes to represent the dependence between related attributes,thus expanding the structure of the Naive Bayesian classifier. Experimental results on the data set show that the KEANBC classifier is superior to the NBC classifier and improves the classification accuracy.