基于Bayes网的软件构件分类

Classifying Software Components Based on Naive-Bayes Network

对软件构件进行分类有助于人们开发高质量的软件。Naive-Bayes网在分类中已经得到成功的应用。但是Naive-Bayes网有一个基本假设:各特征节点要求条件独立。不幸的事,这在现实世界中很难成立。论文利用主成分分析的方法降低了各特征节点的相关性,扩展了Naive-Bayes网的应用范围,并将其用于对软件构件进行分类。实例分析表明新的Bayes分类网预测精度高于一般的Naive-Bayes网。

Classifying software components is helpful to develop high quality software,naive-Bayes network has been used as an effective classifier for many years.However,Naive-Bayes network relies on a basic assumption：the probability distributions for attributes are independent of each other.Unfortunately,it is unrealistic to expect this assumption to hold in the natural worldJn the paper,the author use principle component analysis to relax independence assumption and apply naive-Bayes network into classifying software components,The example given in the paper shows the new one indeed outperforms traditional Naive-Bayes.