Extending developer experience metrics for improving the performance of just-in-time defect prediction models

Abstract

Developers use defect prediction models to efficiently allocate limited quality assurance resources and to plan software quality improvement activities. Traditionally, defect predictions are conducted at the module level. Recently, to overcome the problem of defect prediction at this level, researchers have been actively studying change-level defect prediction, which is known as just-in-time (JIT) defect prediction. When building JIT defect prediction models, researchers use a variety of change metrics that are known to correlate with the bugginess of the changes. Among the widely used change metrics are those that measure the experience of the developer who made the changes. However, these metrics have been found to have no significant impact on the JIT defect prediction models.

In this study, we hypothesized that the use of new developer experience metrics that reflect the characteristics of the projects could improve the performance of JIT defect prediction models and we defined such metrics that reflect the characteristics of the investigated projects. The analysis results of five open-source projects showed that the use of these metrics in JIT defect prediction models is a valid option. Moreover, significant improvements in the cost-effectiveness of these models were found when these metrics were incorporated into them according to the characteristics of the target project.