Article: Unsupervised machine learning for device clustering and dynamic power allocation in hybrid NOMA-MEC system Journal: International Journal of Wireless and Mobile Computing (IJWMC) 2025 Vol.29 No.4 pp.326 - 339 Abstract: Non-Orthogonal Multiple Access (NOMA) and Mobile Edge Computing (MEC) are key 5G technologies aimed at increasing the capacity and efficiency of next-generation wireless networks. However, existing clustering methods face significant challenges, including poor computational efficiency, sub-optimal clustering performance and difficulty in ensuring minimum rate guarantees under dynamic network conditions. To address these issues, this paper proposes an Enhanced <em>K</em>-Means Clustering (EKC) algorithm which dynamically optimises device clustering and power allocation to ensure minimum rate requirements in a hybrid NOMA-MEC system. Results demonstrate that the EKC algorithm surpasses other clustering methods, including Hierarchical, Density-based spatial clustering of applications with noise (DBSCAN) and Gaussian Mixture Model (GMM), in terms of computational efficiency and clustering performance. Theoretical analysis further supports these findings, showing that using Near-Far (NF) pairing as a benchmark, the sum-rate capacity improvements for Quadrature Near-Far (Q-NF), EKC, DBSCAN, Hierarchical and GMM clustering are −0.005%, 9.14%, −47%, 8.23% and 5.77%. Inderscience Publishers - linking academia, business and industry through research

Title: Unsupervised machine learning for device clustering and dynamic power allocation in hybrid NOMA-MEC system

Authors: Sandeep Singh Rana; Gaurav Verma; O.P. Sahu

Addresses: Department of Electronics and Communication Engineering, National Institute of Technology, Kurukshetra, Haryana, India ' Department of Electronics and Communication Engineering, National Institute of Technology, Kurukshetra, Haryana, India ' Department of Electronics and Communication Engineering, National Institute of Technology, Kurukshetra, Haryana, India

Abstract: Non-Orthogonal Multiple Access (NOMA) and Mobile Edge Computing (MEC) are key 5G technologies aimed at increasing the capacity and efficiency of next-generation wireless networks. However, existing clustering methods face significant challenges, including poor computational efficiency, sub-optimal clustering performance and difficulty in ensuring minimum rate guarantees under dynamic network conditions. To address these issues, this paper proposes an Enhanced K-Means Clustering (EKC) algorithm which dynamically optimises device clustering and power allocation to ensure minimum rate requirements in a hybrid NOMA-MEC system. Results demonstrate that the EKC algorithm surpasses other clustering methods, including Hierarchical, Density-based spatial clustering of applications with noise (DBSCAN) and Gaussian Mixture Model (GMM), in terms of computational efficiency and clustering performance. Theoretical analysis further supports these findings, showing that using Near-Far (NF) pairing as a benchmark, the sum-rate capacity improvements for Quadrature Near-Far (Q-NF), EKC, DBSCAN, Hierarchical and GMM clustering are −0.005%, 9.14%, −47%, 8.23% and 5.77%.

Keywords: clustering algorithms; EKC; NOMA; MEC; ML.

DOI: 10.1504/IJWMC.2025.149196

International Journal of Wireless and Mobile Computing, 2025 Vol.29 No.4, pp.326 - 339

Received: 12 Nov 2024
Accepted: 12 May 2025
Published online: 17 Oct 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article

Title: Unsupervised machine learning for device clustering and dynamic power allocation in hybrid NOMA-MEC system

Keep up-to-date