A Support Vector Machine with Elastic Net Regularization and Radial Basis Function based Spectrum Sensing for Cognitive Radio Networks

In recent years, communication technologies are growing significantly and Cognitive Radio (CR) networks is an
expert system to adjust the radio spectrum. However, wireless communication diverse scenarios and distinguishing spectrum occupancy poses a significant challenge in Spectrum Sensing (SS). As it requires high-performance and flexible solutions to accommodate varied characteristics and ensure seamless connectivity. Hence, a Machine Learning (ML) based algorithm namely Support Vector Machine along with Elastic Net Regularization and Radial Basis Function (SVM-ENR-RBF) is proposed to detect and classify spectrum signals. Initially, the spectrum signals are collected from RadioML2016.10b dataset which are preprocessed by Min-Max scaler to normalize Inphase (I) and Quadrature Components (QC) of modulated signals. Finally, SVM classifier provides a regularization technique namely ENR and a kernel function RBF to make easier to analyze as well as classify the spectrum occupancy. The combination of SVM-ENR-RBF improves the detection accuracy, robustness and generalization capabilities. From the results, SVM-ENR-RBF method offers high results of probability of detection, prediction accuracy, and computation time results as 99.8%, 99.2%, and 1.6sec respectively when compared with existing Reinforced Learning-Extreme Learning Machine.