A scalable edge-intelligent spectral clustering framework with random binning features for wireless sensor networks

Wireless Sensor Networks (WSNs) have emerged as a key enabling technology for smart cities, industrial Internet of Things (IIoT), environmental monitoring, and cyber-physical systems, where energy-efficient and scalable clustering remains a significant challenge. Although spectral clustering enables topology-aware cluster formation, its application to large-scale WSNs is constrained by the computational and memory overhead associated with similarity graph construction and spectral embedding. To address this problem, this study aims to develop a scalable and energy-efficient clustering framework that reduces the computational burden of spectral clustering while improving communication performance and network lifetime in large-scale WSNs. To achieve this objective, a scalable edge-intelligent spectral clustering framework based on Random Binning Feature approximation is proposed. The framework integrates randomized feature embedding, sparse graph construction, and partial spectral decomposition to reduce computational and memory overhead while preserving topology-aware clustering. In addition, an edge-intelligent multi-objective cluster-head selection mechanism jointly considers residual energy, predicted residual energy, neighborhood energy distribution, node degree centrality, communication traffic load, and spatial proximity to achieve balanced energy utilization and adaptive communication. The proposed framework is evaluated using NetSim 14.02 under three deployment scenarios comprising 64, 100, and 200 sensor nodes and is compared with LEACH, HEED, K-Means, Adaptive FCM, and ASO–PSO. Experimental results demonstrate superior energy efficiency and throughput performance across the evaluated scenarios. In terms of network lifetime, measured using the Last Node Dead (LND) metric, the proposed framework extends network lifetime by 7.46%, 14.89%, and 2.33% in the 64-, 100-, and 200-node scenarios, respectively, compared with the best-performing benchmark in each scenario. These results demonstrate that the proposed framework effectively improves energy efficiency, communication performance, and network lifetime while maintaining scalability for large-scale WSN and IoT applications.