Supporting change-prone class prediction

Abstract

During the development and maintenance of software, changes can occur due to new features, bug fixes, code refactoring, or technological advancements. In this context, change-prone class prediction can be very useful in guiding the maintenance team, since it is possible to focus efforts on improving the quality of these code snippets and make them more flexible for future changes. In this work, we have proposed a guideline to support the change-prone class prediction problem, which deals with a set of hardworking strategies to improve the quality of the predictive models. Besides, we have proposed two data structures that take the temporal dependencies between these changes into account, called Concatenated and Recurrent approaches. They are also called dynamic approaches, in contrast with the conventional state-of-art static approach. Our experimental results have shown that the proposed dynamic approaches have had a better Area Under the Curve (AUC) over the static approach.