A Novel Scheme for Spectrum Prediction in Cognitive Radio Networks

An efficient spectrum prediction model is presented to improve the spectrum utilization in cognitive radio network. In this model, a novel improved version of Teaching-Learning-Based-Optimization algorithm, also referred to iTLBO algorithm, is proposed to train a feedforward artificial neural network (ANN). The performance of the proposed iTLBO-ANN model is compared with some hybrid prediction models, including the genetic algorithm with ANN (GA-ANN), the firefly algorithm with ANN (FF-ANN), and the conventional TLBO algorithm with ANN (TLBO- ANN). Performance evaluation via a real-word spectrum dataset (GSM-900) confirms that iTLBO-ANN outperforms other spectrum prediction schemes in terms of prediction error and prediction efficiency.