On the prediction of continuous integration build failures using search-based software engineering
Islem Saidani, Ali Ouni, Moataz Chouchen, Mohamed Wiem Mkaouer
Abstract
Continuous Integration (CI) aims at supporting developers in integrating code changes quickly through automated building. However, in such context, the build process is typically time and resource-consuming. As a response, the use of machine learning (ML) techniques has been proposed to cut the expenses of CI build time by predicting its outcome. Nevertheless, the existing ML-based solutions are challenged by problems related mainly to the imbalanced distribution of successful and failed builds. To deal with this issue, we introduce a novel approach based on Multi-Objective Genetic Programming (MOGP) to build a prediction model. Our approach aims at finding the best prediction rules based on two conflicting objective functions to deal with both minority and majority classes. We evaluated our approach on a benchmark of 15,383 builds. The results reveal that our technique outperforms state-of-the-art approaches by providing a better balance between both failed and passed builds.