Cloud computing enables on-demand access to scalable computing resources, fostering agility and cost-effectiveness. However, effective resource allocation strategies are crucial to ensuring optimal utilization and meet dynamic workload demands. Traditional heuristic approaches often lead to suboptimal decisions, performance degradation, and resource wastage in dynamic cloud environments due to multifarious shortcomings. Furthermore, the need for a robust process prediction method is crucial, as doing so would make it possible to schedule said processes before arrival, in turn increasing performance metrics. In order to mitigate these challenges, this research proposes a comprehensive strategy that incorporates machine learning based process prediction and metaheuristic optimization techniques in the process scheduling. The research includes a process prediction pipeline, employing Linear Regression and Long Short Term Memory. However, early predictions would seem futile without efficient scheduling, which is why the article further proposes Global and Local Optimized Preemptive Scheduling (GLOPS), an enhanced metaheuristic scheduling algorithm employing Particle Swarm Optimization and Simulated Annealing. The performance of the proposed work has been evaluated on relevant data, and the results demonstrate how well the method performs, considerably enhancing system performance above conventional heuristics in terms of assessment criteria. The proposed approach positions the method as a competent solution in the field of process prediction and efficient scheduling, all the while outperforming the existing approaches by managing to reduce overall turn around time by up to 48.79%.