基于主动学习的跨项目软件缺陷预测方法

通过软件缺陷预测可以有效地提高软件测试效率，保证软件产品的质量。针对新开发的项目面临训练数据不足，标注代价高以及源项目与目标项目的缺陷模式难以匹配的问题，提出基于主动学习的跨项目软件缺陷预测方法。首先使用主动学习方法对目标项目进行筛选标注，其次将得到的标签集与跨项目数据进行数据融合和模式匹配，最后构建跨项目软件缺陷预测模型。采用真实的软件缺陷数据进行实验，与传统方法比较性能有所提升。结果表明该方法可以通过模式匹配有效提高跨项目软件缺陷预测模型的性能。

An Active Learning-Based Approach to Cross-Project Software Defect PredictioMI Wen-bo1, LI Yong1,2,3*, CHEN You-ren1,2

Software defect prediction techniques can effectively improve software testing efficiency and ensure the quality of software products. For newly developed projects, they face the problems of insufficient training data, high annotation cost, and difficulty matching the defect patterns of source and target projects. This paper proposes a cross-project software defect prediction method based on active learning. Firstly, the target project is filtered and labeled using the active learning method. Secondly, the obtained label set is fused with cross-project data and pattern matching is performed. Finally, a cross-project software defect prediction model is constructed. Real software defect data is used for experiments, and the performance is improved compared with the traditional method. The results show that the method can effectively improve the performance of cross-project software defect prediction models through pattern matching.