The increasing complexity and scale of modern software delivery pipelines have raised the importance of intelligent DevOps management for ensuring system reliability and continuous integration. However, challenges such as noisy time-series data, class imbalance, and fluctuating operational behaviors hinder the effectiveness of traditional monitoring and rule-based automation in dynamic DevOps environments. To address these issues, this paper proposes a predictive framework that utilizes supervised machine learning techniques to forecast system states based on real-time sensor inputs. The proposed methodology integrates rolling window-based feature extraction and normalization, followed by feature selection using Recursive Feature Elimination (RFE) with Random Forests (RF) to isolate the most informative variables. Using time-series data from the HELENA2 dataset, three classifiers, RF, Support Vector Machine (SVM), and XGBoost, were trained and evaluated across multiple performance metrics, including accuracy, precision, recall, and F1-score. Experimental results demonstrate that XGBoost consistently outperformed the other models, achieving an accuracy of 99.1% and an F1-score of 99.25%, indicating superior classification capability. This paper contributes a robust and scalable approach for enhancing DevOps observability through predictive analytics, enabling proactive system management and data-driven decision-making in complex operational environments.
DevOps, Predictive Modeling, Machine Learning, Feature Engineering, XGBoost Classification
IRE Journals:
Ashish Gupta
"A Predictive Framework for Managing DevOps Practices Using Machine Learning Models" Iconic Research And Engineering Journals Volume 4 Issue 9 2021 Page 362-373
IEEE:
Ashish Gupta
"A Predictive Framework for Managing DevOps Practices Using Machine Learning Models" Iconic Research And Engineering Journals, 4(9)