An Innovative Ladybug Beetle Optimized Multi‐Granularity Gated Temporal Convolutional Network for Resource Allocation in 5G Wireless Networks

ABSTRACT The rapid development of 5G communication systems has further increased the popularity of wireless personal communication. However, for improving existing wireless personal communication networks, meeting rigorous broadcast rapidity and QoS requirements is an obstacle. Non‐orthogonal multiple access and massive multiple‐input multiple‐output are two emerging technologies with enormous potential. RA at the BS is crucial to ensure fair distribution of services among users. Increasingly, as the number of users goes up, it becomes indispensable for WNs to elevate their capacity to address increasing demands. One of the significant ways to do this is by the integration of MIMO‐NOMA. This manuscript proposes a novel deep learning‐based resource allocation approach to maximize RA efficiency in MIMO‐NOMA. First of all, the users are grouped. For that purpose, a highly correlated K‐means clustering technique has been introduced, which helps control interference across and within clusters. Further, the aim is to improve the spectral efficiency, maximize the system capacity, and ensure fairness in resource allocation using a Multi‐Granularity Gated Temporal Convolutional Neural Network. Also, the LBO technique has been proposed to optimize the weight parameter of MG‐TCNN to improve the SE performance. A decay‐based movement adjustment was also incorporated into the LBO to maintain exploration and exploitation in better terms for improved convergence behavior. The proposed framework has been administered and simulated via the Python platform. Several assessment metrics, such as ASR, SE, and EE have been computed and associated with other frameworks. A comprehensive ASR of 88.76 Mbps, SE of 119.40 bits/s/Hz, and EE of 19.43 bits/joules/Hz is obtained by the proposed framework for allocating the resources to the 5G WNs.