The method of automatic defect warnings classi cation 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 classification based on trace mining. We use data mining techniques to work on the warning traces to divide them into different groups by the code structure similarity, and make the final warning report easier to manual inspection. Experiments show that this method can classify test results and improve the efficiency of warning inspection in large software testing process.