Predictive Framework for Energy Efficiency in Cloud Computing Using Regression Models and Metaheuristic Optimization Techniques

The prediction of energy efficiency in cloud computing is vital for reducing costs and supporting green energy solutions in large data centers. This study presents a detailed high-technology forecasting plan that employs regression models with meta-heuristic optimization to predict energy performance effectively. The dataset, retrieved from Kaggle, includes a wide range of system parameters such as CPU and memory usage, network traffic, execution time, power consumption, and energy efficiency labels. Analysis reveals that CPU utilization is the most critical predictor of total energy consumption. Regression models were combined with DTR (Decision Tree Regression), GBR (Gradient Boosting Regression), and a stacked ensemble optimized by the Pelican Optimization Algorithm for hyperparameter tuning. The evaluation process included training, validation, and testing phases using MAE, RMSE, R², MARE, and IOA as performance metrics. After training, the GBPO model achieved superior results, with RMSE of 0.053, R² of 0.975, and IOA of 0.984, outperforming traditional GBR and DTR frameworks. Experimental findings confirm that optimization-based methods significantly enhance prediction accuracy and convergence rates. The practical implications extend to energy-efficient task scheduling, virtual machine consolidation, and power-aware orchestration, enabling cloud providers to minimize energy costs and carbon emissions.