Towards intelligent user clustering techniques for non-orthogonal multiple access: a survey

EURASIP Journal on Wireless Communications and Networking

Table 2 Unsupervised machine learning-based user clustering schemes in NOMA

Clustering scheme	Related work	Scenario	Assumptions	Objective functions	Application	Findings
K-means	[31, 32]	Downlink	User distribution follows the Poison Cluster Process inside cell	Sum rate maximization	mm-wave MISO, UAV	(1) Jointly user clustering, power allocation and beamforming for scheduling of UAV trajectory (2) Optimal user clustering
Fuzzy C-means	[33, 34]	Downlink	CSI is known by the base station	Maximize energy and power	Massive MIMO	Clustering is based on QOS requirement of user and provides fast convergence rate
Enhanced K-means	[35, 36]	Downlink	CSI is known by the base station	Maximize energy efficiency	Tera-Hertz MIMO, MISO UAV	(1) Cache-enabled system to handle heterogeneous environment with fast converging rate (2) Joint user clustering and beamforming for scheduling of UAV trajectory
Hierarchical	[37]	Downlink	User distribution follows the poison cluster process inside cell, perfect CSI is known by the base station	Sum rate maximization	mm-wave MISO	The clustering scheme provides no need to fix the number of clusters and provide more accurate result in case of random user distribution
DBSCAN	[30]	Downlink	User distribution follows the poison cluster process	Sum rate maximization	mm-wave MISO	The clustering scheme provides the QOS-based beamforming