Optimized neural network for spectrum prediction using genetic algorithm in cognitive radio networks

Abstract

Cognitive radio based dynamic spectrum access mechanism is a revolution in wireless communication which alleviates the spectrum utilization problem. Here cognitive users uses spectrum sensing techniques, to sense the bands before transmitting on them to avoid collision with the licensed users which leads to delay and conserves more energy. To reduce delay and energy consumption and to predict the future usage of channels spectrum prediction methodologies used. Spectrum prediction used to predict the future channel status based on collected historical data, here to solve the problem, the neural network based spectrum prediction using back propagation training model has been proposed. To optimize the structure of the neural network and to reduce the aggressive weight structural pattern genetic algorithm is used here to avoid trapping in local optimal solution. Selection, crossover and mutation functions are performed to increase the randomness, which extends the population converge to the set that contains the global optimal solution. Simulation results show the genetic algorithm which is used has increased the probability of obtaining the best weights by optimizing the network, also the proposed scheme results indicate high prediction accuracy.