A Comprehensive Review of Machine Learning Approaches for Predictive Resource Management in Cloud Computing Environment

Abstract

The rapid growth of cloud computing has significantly increased the complexity of managing heterogeneous and dynamic computing resources while ensuring quality of service (QoS), cost efficiency, and scalability. Traditional rule-based and reactive resource management strategies often fail to adapt to highly fluctuating workloads and diverse application demands. Consequently, machine learning (ML) has emerged as a promising paradigm for predictive resource management by enabling intelligent forecasting, proactive decision-making, and automated control of cloud resources. This paper presents a comprehensive review of machine learning approaches employed for predictive resource management in cloud computing environments. It systematically examines supervised, unsupervised, and reinforcement learning techniques, including regression models, support vector machines, decision trees, ensemble learning, deep learning, and deep reinforcement learning frameworks. The review further analyzes their applications in workload prediction, virtual machine allocation, autoscaling, load balancing, energy optimization, and service-level agreement (SLA) violation mitigation. Performance metrics, evaluation datasets, and experimental platforms commonly used in existing studies are also discussed. Additionally, this paper highlights key challengesKey such as data heterogeneity, model interpretability, scalability, and real-time adaptability, and outlines open research issues to guide future investigations. By consolidating current knowledge and identifying research gaps, this review aims to assist researchers and practitioners in developing efficient, reliable, and intelligent cloud resource management solutions.