In cloud computing, Infrastructure as a Service (IaaS) presents task scheduling as an NP-hard problem without polynomial-time solutions. So, using metaheuristic algorithm methods to improve task scheduling, especially swarm intelligence and more specifically beluga whale optimization and its fitness function, is seen as very promising. By incorporating predictive analysis and machine learning techniques into this algorithm, researchers can enhance its accuracy, comprehend the critical nature of task scheduling, and enhance cloud load balancing, leading to improved resource allocation and utilization. The novelty of using predictive analysis will also help to understand the complex nature of task scheduling in a cloud computing environment. Using machine learning for predictive analysis with the neural network method in beluga whale optimization (BWO) meta-heuristic search algorithm will allow us to look in- to the cloud data center load balancing problem in new ways. This research article’s proposed algorithms and their results have been tested and run in a specially created simulation environment under Python 3.12 version. The proposed approach incorporates neural networks to enhance predictive accuracy, optimize resource allocation, and improve load balancing. Results demonstrate a significant performance advantage over traditional optimization techniques, with an R-squared value of 0.83, confirming 83% predictive accuracy. This work contributes to cloud computing by offering a scalable, energy-efficient task scheduling framework that leverages historical data for real-time decision-making.