| The acceleration of social informatization has led to an increase in the demand for wireless mobile services,which in turn has led to a significant increase in the demand for spectrum resources.However,the existing spectrum resources and their allocation methods cannot meet the demand for frequency usage,which has limited the development of wireless technology to some extent.In 1999,Cognitive Radio technology was proposed,which uses spectrum awareness and intelligent learning of the system to achieve dynamic allocation and sharing of spectrum,thus enabling unauthorized users(Secondary Users)can use the corresponding frequency bands without affecting the authorized users.Spectrum forecasting is the process of analyzing historical data on band usage and using these data to predict future spectrum usage.Although traditional spectrum prediction models can provide a certain degree of accuracy,they are susceptible to the influence of local optima.To address these drawbacks,this paper introduces a swarm intelligence optimization algorithm to optimize the neural network model to improve the prediction accuracy.The main work carried out in this paper is as follows:(1)The traditional Bald eagle search optimization algorithm is improved.Firstly,inertial weights are used in the search phase of the algorithm to achieve the position update.Secondly,the golden sine algorithm is incorporated to capture the prey.Finally,the vertical and horizontal crossover strategy is used as a way to achieve the purpose of maintaining the convergence speed while jumping out of the local optimum,so as to improve the overall search capability.The improved algorithm has been shown to have a remarkable enhancement in convergence accuracy through experimental results.(2)To address the problem that noise may exist in the original data and the noise directly affects the model prediction results,the ensemble empirical modal decomposition method is introduced to decompose the original data and overcome the modal confusion problem in signal decomposition;secondly,the sample entropy method is introduced to reconstruct the subseries according to the similar modal values obtained from the decomposition.The experimental results demonstrate that the proposed method can improve the prediction accuracy.(3)The improved intelligent algorithm is used to optimize the weights and thresholds of the BP neural network model to improve the prediction accuracy and precision of the model;The decomposed-integrated data are used in the BP neural network model and compared with the unprocessed data and the decomposed data only,the simulation results prove that the proposed hybrid model can improve the accuracy of prediction. |
