基于踪迹挖掘的自动缺陷警报分类方法

缺陷检测一般包括静态分析与人工确认两个阶段.静态缺陷检测工具报告大量警报,但是主要的警报确认工作仍然由人工完成,这是一件费时费力的工作.巨大的确认投入,会导致测试人员和管理人员拒绝使用该静态检测工具.为了辅助警报确认工作,提出一种基于警报踪迹挖掘的警报分类方法,使用该方法挖掘警报踪迹进而将代码结构相似警报分为一类,使得分类后的最终警报报告更加易于人工确认.实验表明,该方法能够在较大规模的软件测试过程中分类测试结果,提高警报确认效率.

The method of automatic defect warnings classification based on trace mining

Defect detection generally includes static analysis and warning inspection two stages. A large number of defect warnings reported may lead developers and managers to reject the use of static analysis tools as part of the development process due to the overhead of warning inspection. To help with the inspection tasks,proposed a method of automatic defect alarms classifcation based on trace mining. We use data mining techniques to work on the warning traces to divide them into diferent groups by the code structure similarity,and make the fnal warning report easier to manual inspection. Experiments show that this method can classify test results and improve the efciency of warning inspection in large software testing process.