Efficient task scheduling is the most important task in cloud computing system for the optimal performance

optimization. Traditional informed searches like the First-Come-First-Served (FCFS) and Round Robin often fail to account for the dynamic nature of the modern-day tasks and workloads, which increases the latency and fails to utilize resource properly. Our paper proposes a Machine learning (ML) driven approach for the correct and efficient prediction of task waiting time and schedule the tasks intelligently. We have used a public cloud workload dataset which includes 5000 job entries, and have implemented a Random Forest regression model to analyze the impact of multi-dimensional features like CPU utilization, memory consumption and network bandwidth. The results demonstrate that Machine Learning models can effectively identify the performance matrices and the network bandwidth has emerged as the primary predictor of task latency. Our proposed framework provides a foundation for smart schedulers which can dynamically allocate the resources which reduces the error rates significantly and also enhances the system throughput.