基于自学习的软件质量实时预警模型

SOFTWARE QUALITY REAL-TIME EARLY WARNING MODEL BASED ON SELF-LEARNING

软件质量的管理和控制在软件开发项目管理研究中十分重要,但现有的软件质量分析模型通常都是针对软件最终产品进行评估,并且其对于数据集中产生的变化不敏感,难以适应多变的开发环境。利用主成成分分析法(PCA)和一种可自主生成学习样本集的增量支持向量机(IL-SVM)提出一种可以自我进行规则增量学习的实时软件质量进行分析的预警模型。通过主成成分分析法来降低原始数据集的属性维度,然后在利用可增量学习的支持向量机来进行软件质量预警,最后利用NASA的数据集进行相关实验。结合实际情况进行分析,该方法在质量预警方面具有较好的效果。

The management and control of software quality are very important in the research of software development project management. However,the existing software quality analysis models are usually evaluated for software end products and are insensitive to changes in data sets and difficult to adapt to ever-changing development environment. By using the method of principal component analysis( PCA) and a self-generating incremental learning support vector machine( IL-SVM),we proposed a self-warning model of real-time software quality analysis model. The principal component analysis( PCA) was used to reduce the attribute dimension of the original dataset,and then the software quality warning was carried out by using the incremental learning SVM. Finally,the relevant experiments were carried out by using the NASA dataset. Combined with the actual situation of the analysis,the method had good results in quality warning.