An Intelligent Multi-Class XGBoost-Based Model for Optimizing DevOps Continuous Integration and Continuous Deployment Failure Prediction

Modern software development fundamentally relies on agile methodologies and DevOps practices to facilitate accelerated software delivery. Continuous integration and continuous deployment CI/CD are among the most critical DevOps practices that require considerable attention to execute successfully. Therefore, this study proposes a multi-class XGBoost-based model to improve the performance of CI/CD failure prediction. The proposed model was trained and tested using the comprehensive TravisTorrent dataset, which contains extensive build information from several projects developed in various programming languages. The experimental results demonstrate that the proposed model achieves a statistically significant performance improvement of nearly 18% over SVM and the Random Forest models. Beyond the performance improvement, SHAP (SHapley Additive exPlanations) analysis was employed to explain the model’s decision-making process, revealing that the most influential features, ranked in descending order of importance, are build log status, build duration, build start time, the number of commits in the repository, and repository age. This interpretability enhances both the reliability and transparency of the proposed model.