Towards the Efficient Management of Cloud Resource Allocation: A Framework Based on Machine Learning

Journal of Artificial Intelligence and Big Data | Vol 2, Issue 1

Table 1. Overview of Recent Studies on Cloud ResourceAllocation Using Machine Learning

Author	Proposed Work	Dataset	Key Findings	Challenges/recommendation
Chudasama and Bhavsar (2020)	DL + Queuing Theory model for proactive auto-scaling	University server logs	Improved SLA violation prediction by 5%, Enhances resource elasticity under hybrid cloud	Static threshold auto-scaling fails under unpredictable loads, need for proactive, prediction-driven auto-scaling mechanisms in hybrid cloud environments
Chen et al. (2019)	A self-adaptive system for allocating resources for cloud-based software applications and self-learning, utilizing genetic algorithms for optimization and machine learning for QoS modelling.	RUBiS benchmark	QoS prediction accuracy > 90% 10%–30% improvement in resource utilization	Traditional policy-driven methods lead to complexity and high administrative cost; recommends ML-driven automatic decision-making to adapt to dynamic environments.
Rayan and Nah (2018)	ML-based workload prediction for cloud data centers (RFR, SVR, PR)	Operational workload logs	RFR achieved lowest RMSE (11.68 for PMs, 4869.08 for PC), 2-second training time Enables proactive allocation and energy/resource efficiency	Focused on prediction, not dynamic real-time scheduling, Need to integrate accurate workload prediction with adaptive scheduling/auto-scaling mechanisms in large-scale environments
Ataie et al. (2017)	Hybrid methodology that integrates support vector regression (SVR) and queuing networks to forecast the duration of job execution	Hadoop MapReduce job traces	Achieved 21% improvement in prediction accuracy over standalone ML methods	Need to balance accuracy and computational cost, Integration of analytical models and ML recommended for better resource management
Dai et al. (2016)	A method for multi-objective optimization that is intended to maximize the price, accessibility, and efficiency of cloud-based Big Data programs. carried out on the testbed.	Experimental setup	Execution time improved by 20% over traditional methods- 15% higher performance than heuristics- 4–20% cost savings	Emphasizes the need for fine-grained resource allocation in cloud infrastructure; recommends multi-objective optimization to handle competing objectives.