Cooperative Spectrum Occupancy Based Spectrum Prediction Modeling

Spectrum sensing is the cornerstone of successful deployment of cognitive radio technology. Accurate models need to be developed to enhance the accuracy of spectrum sensing thereby reducing interference to the primary users. Neural network based spectrum prediction has the ability to learn from historical data and doesn’t need to be built from scratch every time the model needs to be updated. These models need to be accurate and reliable. In this paper, an attempt was made to improve the prediction accuracy and reliability of the model by using Genetic Algorithm to optimize the weights. The data used was from a cooperative spectrum sensing which involves two devices, unlike the normal spectrum occupancy that involves a single device; this improves the reliability of the prediction as the probability of false alarm has been reduced. Two metrics were used to ascertain the performance of the model; they both indicate an improvement over a normal neural network model.