摘要
采集软件研发过程中可能与缺陷有关的过程数据或产品数据,对软件缺陷数量进行预测,达到对软件质量的把控。采用LASSO进行特征值选择确定最佳影响因子集合,采用线性模型和贝叶斯网络模型分别对样本数据进行预测,说明两种模型的因子分析过程和模型构建过程,采用R语言进行编码实现。通过预测结果的对比验证了当数据经过二次主观加工后,采用线性模型的预测结果比贝叶斯网络预测结果更准确。
According to the collection of processing and product data of software development relating to the defect,it can make a prediction of the number of the software bug to control the software quality.The best impact cues assemblage was confirmed using LASSO characteristic method,the number of bug was predicted using two methods including linearity method and Bayes network method based on the sample data.The processes of development were described using these two methods and R language was used to achieve encoding.Comparing the prediction results of two methods,it verifies that the linearity method is more accurate than the Bayes network method because of the second objective operation on sample data.
作者
马由
汤艳
解斐
MA You;TANG Yan;XIE Fei(Software Testing Center,China Electronics Technology Group Corporation No.15 Research Institute,Beijing 100083,China)
出处
《计算机工程与设计》
北大核心
2020年第4期1046-1051,共6页
Computer Engineering and Design
