CLOUD COST-PERFORMANCE MODEL ANALYSIS AND OPTIMIZATION USING MACHINE LEARNING

Abstract

Cloud computing has revolutionized modern computing by providing scalable, on-demand resources, but optimizing the trade-off between cost and performance remains a significant challenge. This research addresses the gap in understanding and predicting cloud cost-performance relationships by introducing a machine learning framework. The primary objective is to enhance decision-making for resource allocation in diverse cloud environments, leveraging data-driven insights. The proposed approach employs a Feature-Aware Gradient Boosting (FA-GB) model, which dynamically adjusts learning rates based on feature importance, enabling a balanced learning process and improved prediction accuracy. This novel method effectively handles complexities such as workload variability and data imbalance, achieving superior predictive performance. Algorithms such as Gradient Boosting and feature-specific adjustments are integrated into the framework to optimize execution time, energy efficiency, CPU usage, and memory usage. Experimental results demonstrate the model’s ability to achieve low error rates, with competitive Mean Squared Error (MSE) and Mean Absolute Error (MAE) values across all metrics. The integration of predictions into a composite performance score enables a holistic evaluation, revealing critical trade-offs and guiding resource optimization strategies. The findings underscore the significance of data-driven methods in optimizing cloud systems, offering actionable insights to improve efficiency, reduce operational costs, and enhance system reliability. This research advances sustainable cloud infrastructure by addressing dynamic, heterogeneous environments with an adaptive, high-performance predictive framework.