摘要
为了确切地估计软件缺陷分布,提出了基于AODE和再抽样的软件缺陷预测模型。分析了几种常用贝叶斯分类器的优缺点,以及软件缺陷度量元和再抽样方法,设计了多AODE分类器,该分类器是由多个AODE二分类器组成的。在以上基础上,建立了采用多AODE分类器和再抽样方法的软件缺陷预测模型,实验结果表明,该分类器的分类准确度较常用贝叶斯分类器高。通过收集到的缺陷数据样本比较结果表明,该模型比一些常用贝叶斯模型具有更好的预测准确性和稳定性。
To evaluate software defect distribution exactly,a software defect prediction model based on AODE and resampling is put forward.Firstly,some common Bayesian classifiers are introduced briefly,and their advantages and deficiencies are pointed out.Then, software defect metric units and resampling methods are analyzed,and multi-AODE classifier formed by multiple AODE 2-classifier is designed.Comparative experiments show that multi-AODE classifier has better classification accuracy performance than some common Bayesian classifier.On this basis,the software defect prediction model based on multi-AODE classifier and resampling is constructed. Finally,collected defect data samples are used to verify the model and the result indicated that the model is better than the common Bayesian models both in veracity and stability of prediction.
出处
《计算机工程与设计》
CSCD
北大核心
2011年第1期210-212,300,共4页
Computer Engineering and Design
关键词
平均单一相关评估器
贝叶斯分类器
软件缺陷度量
再抽样
软件缺陷预测
aggregating one-dependence estimators
Bayesian classifier
software defectmetric
resampling
software defect prediction