Enhancing spectrum prediction in cognitive radio networks using an optimized generative adversarial network
G. Narmadha, M. S. Jeyalakshmi, M Ponnrajakumari, Duraichi Natarajan, B. Sakthivel · Results in Engineering · 2025
This work proposes a novel approach for spectrum prediction in Cognitive Radio Networks (CRN) using a Generative Adversarial Network (GAN) model. Most existing spectrum prediction approaches use statistical techniques for spectrum prediction. However, the accuracy of the existing prediction models is very low and does not handle temporal dependencies of data effectively. The proposed GAN model uses a Bidirectional Long Short-Term Memory (BiLSTM) network as the generator and an Echo State Network (ESN) as the discriminator for effective spectrum data processing. The generator generates synthesized spectrum data based on the past data of available spectrum. The discriminator distinguishes between real and generated data. By iteratively training the generator and discriminator, the proposed model aims to accurately predict the available spectrum in CRNs. Also, the hyperparameters of the GAN model are tuned using the optimization algorithm of Red Panda Optimization (RPO) algorithm. Experimental results on real-time data sets show that the proposed GAN-based prediction model improves spectrum prediction accuracy and allows efficient spectrum utilization in dynamic and heterogeneous CRN environments.