Utilizing Azure Automated Machine Learning and XGBoost for Predicting Cloud Resource Utilization in Enterprise Environments
Mahi Ratan Reddy Deva, Nilesh Jain
Abstract
The most popular way for companies of all sizes to satisfy their IT needs is now via cloud computing. The modern computational technology system operates as a fundamental framework which optimizes resource management systems alongside optimization outputs. Many discussions have emerged about integrating Machine Learning (ML) into cloud resources during the recent period. A publicly accessible Kaggle cloud computing performance dataset helps this research explore how ML optimization affects energy consumption alongside performance efficiency. The dataset includes variables such as CPU use, memory usage, power consumption, and task information. The regression models XGBoost Regressor delivered better outcomes than Gradient Boosting Regressor since it used standardized features to reach optimal performance ratings of MAE (0.0979), RMSE (0.1134), <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{R}^{\mathbf{2}}$</tex> (0.9999) and Max Error (0.2977). The deployment of Azure AutoML enabled automatic selection of the best model which was validated by 5-fold cross-validation. XGBoost demonstrates better predictive power than other models, such as Support Vector Regression (SVR) and LSTM, because its evaluation metric values demonstrate superior performance ratings. Visualizations of residuals, pair plots and task-type distributions provided additional support for the conducted statistical analysis. Research results demonstrated that XGBoost provided superior performance to all other models for forecasting CPU usage in cloud environments, thus establishing itself as a powerful tool for sustainable cloud infrastructure management.