Advancing QoS in 5G Wireless Sensor Network Slicing: A Hybrid Approach Using Sparse Spectra Graph Convolutional and Supervised Bidirectional Long Short‐Term Memory Network

ABSTRACT The fast evolution of 5G and next‐generation networks requires immediate network processing combined with high reliability features and effective resource management to enable autonomous systems along with healthcare capabilities and industrial IoT applications. The main structural component of 5G architecture uses network slicing to create exclusive virtual networks that meet the precise quality‐of‐service (QOS) demands of different applications. Existing approaches fail to maintain precise and complete resource allocation with slice management in changing conditions that result in poor operational outcomes. This paper introduces a hybrid approach using sparse spectra graph convolutional networks (SSGCN) and supervised bidirectional long short‐term memory network (SBLSMN) to resolve current challenges in wireless sensor networks (5GWSN‐SGC‐SLMN). The SSGCN optimizes spatial resource allocation by using spectral graph convolutions to understand wireless sensor network (WSN) node interactions during slice selection and the SBLSMN tracks load data through bidirectional learning for error rate prediction to implement proactive QoS management, as the spatial and temporal learning work together to optimize resource allocation and provide flexible network expansion between different network slices. Experimental results show that the proposed approach attains 19.11% better accuracy, 17.12% enhanced precision, and 18.51% reduced misclassification rate than existing methods. The proposed method attains a 32% decrease in URLLC slice latency and delivers 98% reliability because of its ability to monitor real‐time traffic and automatic failure response before slice failures occur. These improvements highlight the effectiveness of 5GWSN‐SGC‐SLMN in ensuring reliable and efficient wireless network slicing in 5G environments.