Cloud computing has witnessed widespread adoption, driven by reduced hardware, software, and maintenance costs. It offers access to a shared pool of virtual resources in energy-intensive data centers, where diverse services have dynamic workloads. These data centers housing large volumes of data and numerous servers face significant energy consumption challenges, resulting in high operational costs and carbon emissions. To mitigate energy consumption, it is critical to prioritize accurate workload prediction. These predictions provide a precise estimate of the energy consumed by data centers and improve scaling for cloud service providers and high-quality service for cloud consumers. This paper presents a comprehensive review of the approaches utilizing Machine Learning (ML) for workload prediction and energy management in cloud computing to optimize resource allocation schemes. It begins with an overview of cloud computing, resource management, and the associated challenges of predicting workloads and energy management. Through a structured comparative analysis, the work discusses various ML approaches and compares them based on their datasets, evaluation metrics, and computational complexities, and providing valuable insights into their strengths and weaknesses, along with their real-world applications. In addition, it also identifies significant research gaps and promising avenues for the development of adaptive and energy-efficient cloud resource management. The research findings will assist researchers and practitioners in selecting appropriate ML models while encouraging the development of more sustainable and scalable cloud infrastructures.