What type of study is this?

This is a Experimental Study study.

October 9, 2025Open Access

A Novel Load Balancing Method by Integrating Machine Learning Task Classification With Federated Learning and Hybrid Meta-Heuristic Optimization

Key Points

Efficient load balancing is achieved by integrating machine learning for task classification and federated learning principles.
The method manages discrete task types—audio, text, image, and video—to optimize the allocation of resources effectively.
Hybrid optimization methods, including Tasmanian Devil Optimization and Beluga Whale Optimizer, enhance workload distribution for better performance.
The algorithm preserves privacy by relying on federated learning, localizing data while sharing only model updates, thereby improving system resilience.

Abstract

In computer computing, the efficient load balancing of a system is required to optimize the utilization of resources and enhance the performance of a system. This paper proposes a novel method of task classification using machine learning, integrating it with federated learning principles and a hybrid meta-heuristic optimization method, which is efficient in load balancing. The very first thing is done by categorizing the tasks into discrete types: audio, text, image, and video. This is done using state-of-the-art machine learning methods. Such classification helps understand computing requirements for each class of tasks; hence, precise resource allocation can be facilitated. Following this, the categorized tasks undergo a hybrid process of optimization with Tasmanian Devil Optimization and the Beluga Whale Optimizer. These techniques are used to achieve the fair distribution of workloads over a network of computers. The algorithm has a levy flight strategy, thus can effectively avoid local optima and speed up the convergence towards the optimal solutions. Federated learning has improved the system's ability to handle distributed and decentralized data sources without compromising privacy. In addition, federated learning will enable cooperative model training between several devices while keeping data localized. This ensures that there are no privacy concerns as the generalization abilities of the load-balancing model are increased. Data collection under federated learning is done in a way that keeps data on local devices, only sharing model updates to central servers. The simulation showed that this approach improves the distribution of workloads and system resilience, thus providing a robust framework for handling complex fluctuating computing environments. It has been implemented in Python, the evaluation has been made over the state-of-the-art approaches in terms of throughput, energy consumption, and makespan as well. The proposed approach optimizes not only the resource allocation but also the overall efficiency and scalability of the system.

Read Full Paperexternally

Demander à l'IA

Bookmark

View Full Paper