<p>Managing resources in data centers remains a critical challenge due to the growing complexity of infrastructure and the dynamic nature of workloads. The accurate estimation of future resource demand and the utilization of these outcomes to schedule resources to the Virtual Machine (VM) is crucial for optimizing resource utilization in a cloud environment. Current mechanisms focus either on forecasting processes or implementing scheduling methods, but very few papers explore the collaboration of these two techniques. Furthermore, they forecast through a random selection of resources without considering significant features. It substantially impacts the effectiveness of intelligent decision-making, resource utilization, energy consumption, and maintenance costs for hosts. This paper presents a methodology grounded on a prediction-based scheduling approach. It involves several factors such as analyzing best-fit features, forecasting the upcoming workload using a Temporal Dynamic Ensemble Approach (TDEA), evaluating the algorithms with performance parameters, and scheduling resources based on the forecasted output using the Whale Optimization Algorithm with Pareto Front. To show the feasibility of the proposed work, this methodology is applied to realistic workload datasets obtained from various data centers, as well as on a synthetic dataset. The forecasting model for the upcoming 60 instances exhibited the most favorable Root Mean Square Error (RMSE) value of 0.1437, outperforming baseline models like LSTM and GRU as much as 22.6%. Additionally, the scheduler considers these expected instances as input, and then the scheduler distributes them among the virtual machine cores. This strategy aims to achieve optimal outcomes by aligning resource allocation with the system's changing demands. The experimental results demonstrate that the scheduling phase reduces the overall completion time by 20.3%, in energy consumption by 17.8%, and by 14.9% in cost when compared to existing approaches.</p>