Hybrid Deep Learning-based Load Balancing Model using Feature Engineering with Gradient Boosting Model for Cloud Computing

In cloud computing, load balancing, particularly when using deep learning, is the deliberate distribution of workloads across several resources to maximize system stability, performance, and resource consumption. By judiciously assigning resources to incoming requests, deep learning models can forecast server loads, optimize job scheduling, and boost overall cloud performance. Traditional Load balancing techniques struggle with dynamic and unpredictable workloads, leading to inefficient resource utilization. They often relay on static rules, lacking adaptability and real-time decision-making capabilities. This novel load balancing approach integrates an Extreme Gradient Boosting (XGBoost) with a hybrid optimization framework combining Rock Hyrax Optimization (RHO) and Adaptive Genetic Ant Colony Optimization (GACO). The Extreme Gradient Boosting module predicts the execution time of tasks based on real-time system metrics and historical workload data, enabling intelligent scheduling decisions. Rock Hyrax Optimization is employed for its superior exploration capabilities in dynamic cloud environments, while GACO enhances solution exploitation by combining pheromone-based learning with genetic refinement strategies. These components dynamically allocate tasks to the most suitable virtual machines, reducing system overload and ensuring balanced load distribution. Simulation results demonstrate improved task completion rates, reduced response time, and increased overall throughput compared to traditional load balancing methods.