Resource Allocation in NOMA Networks: Convex Optimization and Stacking Ensemble Machine Learning

• Published in: IEEE open journal of the Communications Society Vol. 5; pp. 5276 - 5288
• Main Authors: Ghanbarzadeh, Vali, Zahabi, Mohammadreza, Amiriara, Hamid, Jafari, Farahnaz, Kaddoum, Georges
• Format: Journal Article
• Language: English
• Published: New York IEEE 2024; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Adaptive systems; Algorithms; Artificial neural networks; Computational efficiency; Convex analysis; convex optimization (CO); Convexity; joint power allocation and channel assignment (JPACA); Machine learning; machine learning (ML); Neural networks; NOMA; non-orthogonal multiple access (NOMA); Nonorthogonal multiple access; Optimization; Quality of service; quality of service (QoS); Quality of service architectures; Real time; Resource allocation; Resource management; Stacking; stacking ensemble method; Uplink; Wireless communication systems; Wireless networks
• ISSN: 2644-125X; 2644-125X
• DOI: 10.1109/OJCOMS.2024.3450207

This article addresses the joint power allocation and channel assignment (JPACA) problem in uplink non-orthogonal multiple access (NOMA) networks, an essential consideration for enhancing the performance of wireless communication systems. We introduce a novel methodology that integrates convex optimization (CO) and machine learning (ML) techniques to optimize resource allocation efficiently and effectively. Initially, we develop a CO-based algorithm that employs an alternating optimization strategy to iteratively solve for channel and power allocation, ensuring quality of service (QoS) while maximizing the system's sum-rate. To overcome the inherent challenges of real-time application due to computational complexity, we further propose a ML-based approach that utilizes a stacking ensemble model combining convolutional neural network (CNN), feed-forward neural network (FNN), and random forest (RF). This model is trained on a dataset generated via the CO algorithm to predict optimal resource allocation in real-time scenarios. Simulation results demonstrate that our proposed methods not only reduce the computational load significantly but also maintain high system performance, closely approximating the results of more computationally intensive exhaustive search methods. The dual approach presented not only enhances computational efficiency but also aligns with the evolving demands of future wireless networks, marking a significant step towards intelligent and adaptive resource management in NOMA systems.