Random Forest Model for Software Build Time Prediction on CI/CD Pipeline

In the fast-paced world of software engineering, Continuous Integration/Continuous Delivery (CI/CD) pipelines are essential to deliver software builds continuously. However, the varying time taken for software builds to complete on these pipelines can challenge scheduling software delivery and impact productivity. To the best of researchers' knowledge, machine learning techniques have never been used to predict software build time in the CI/CD pipeline. This research attempted to apply data science and machine learning techniques, including linear regression (LR), support vector regressor (SVR), random forest regressor (RFR), and XGBoost regressor, to predict software build completion time to address this research gap. Past build events were used as a dataset to train and identify the best-performing model by evaluating the time a software build takes to complete. Different factors contributing to software build time on the CI/CD pipeline were also analyzed to identify opportunities for improvement. This research found that the random forest (RF) model achieved the best and outstanding performance of 14.306 in mean squared error (MSE). This model could be deployed to provide completion time estimates for software builds, enabling better code delivery scheduling. This research also suggested opportunities for improvement in the CI/CD pipeline by discovering major factors causing high build time in the CI/CD pipeline that engineers could rectify to reduce software build time in the CI/CD pipeline.