Machine Learning-Driven CI/CD Pipeline Optimization Using a Support Vector Machine and XGBoost Ensemble
Abstract
CI/CD pipelines are essential to the contemporary software development, yet usually have such flaws as long build times, high failure rates, and poor use of resources. These problems are addressed with the help of predictive analytics and optimization of resources through the application of machine learning (ML). The proposed paper is an amalgam of Support Vector Machine (SVM) and XGBoost to enhance the reliability of the builds, failure prediction, prioritization of tests, and resource prediction of CI/CD pipelines. The model is experimented against publicly available datasets and CI/CD simulation environment, and thus, it is robust and applicable. Comparing the hybrid model to the baseline models, it has a high ROC-AUC and F1-score of 0.96 and 0.91 respectively in predicting build outcomes, and reducing mean time to detection and recovery and raising pipeline throughput by 34%. In order to offer developers empirical operational advantages and explainable results, this paper builds on the margin-based robustness of SVM and nonlinear feature learning of XGBoost. This model is a state-of-the-art extension of the CI/CD stream optimization, which would provide a scalable and interpretable framework that can be deployed across a broad spectrum of software engineering processes.