Hybrid SVM-RDA-based joint optimization of detector threshold and secondary users for energy-efficient cooperating spectrum sensing in fading environment

Abstract The performance of Cognitive Wireless Sensor Networks (CWSNs) can be severely harmed by issues including fading, shadowing, and hidden terminals. Although Cooperative Spectrum Sensing (CSS) shows promise in addressing these problems, it comes with significant costs, especially in terms of energy usage. To improve the overall performance of the system, battery-powered secondary users (SUs) operating in unlicensed frequency bands must optimize their energy efficiency (EE). Nevertheless, prior research frequently ignores the impacts of fading and noisy reporting when analyzing CSS performance. This study bridges the gap by highlighting EE as a critical performance metric that carefully analyzes CSS under actual channel conditions. We offer a novel Hybrid Support Vector Machine-Red Deer Algorithm (SVM-RDA) and provide a system model that evaluates EE in centralized CSS over the Nakagami fading channel in order to jointly optimise the total number of SUs and the detector threshold. The results demonstrate that the proposed SVM-RDA strategy outperforms traditional optimization techniques as the Whale Optimisation Algorithm (WOA) and Sub Optimal Iterative Search Algorithm (SOISA) in terms of EE. Specifically, the SVM-RDA finds that the ideal number of SUs is 16, the ideal detector threshold is 13.10, and the maximum EE is 3.28 Mbits/Hz/Joule. The findings significantly improve EE and CSS for CWSNs and highlight the necessity of collaboratively adjusting the scheme design to achieve a balance between energy economy and detection accuracy.