The increasing complexity of tasks on mobile devices has escalated energy consumption, impacting both user experience and the environment. This review focuses on optimizing energy efficiency in mobile edge computing (MEC) by leveraging machine learning (ML) techniques, particularly Deep Reinforcement Learning (DRL). MEC aims to offload computational tasks to the network's edge, thereby reducing the energy burden on mobile devices by utilizing nearby edge resources. This comprehensive review addresses the gap in the literature by extensively comparing and contrasting various ML-based approaches for energy optimization in cloud computing. The study evaluates different ML models, tools, datasets, and metrics, with a specific focus on energy efficiency. Our analysis indicates that DRL, due to its ability to handle complex environments and learn from interactions, is the most widely adopted model, featuring in 55% of the related works reviewed. Additionally, TensorFlow emerges as the predominant tool, used in 35% of the studies, thanks to its flexibility, scalability, and robust community support. The AudioSet dataset is frequently utilized, accounting for 28% of the reviewed works, due to its compatibility with deep learning frameworks. The primary objective across these studies is to reduce energy consumption, driven by the need to prolong battery life, enhance usability, and mitigate environmental impact. The most common limitation identified is high computational demand, present in 15% of the studies. This review concludes that DRL, supported by TensorFlow and the AudioSet dataset, presents the most promising avenue for future research in energy-efficient mobile edge computing.