Network traffic prediction based on RBF neural network optimized by improved gravitation search algorithm

Abstract

In order to improve the setting method of the parameter and network structure of neural network, this paper takes neural networks and chaotic time series theory as the foundation to propose a prediction model based on RBF neural network optimized by improved gravitation search algorithm. Against the disadvantages of gravitation search algorithm such as slow convergence and prone to premature convergence, the improved gravitation search algorithm improves the gravitational coefficient and speed selection formula and chooses the positions of updated particles by the survival of the fittest selection law, which can better balance local and global search capabilities. We apply the RBF neural network optimized by improved gravitation search algorithm to Lorenz chaotic time series and network traffic to test the validation of the algorithm and compare with RBF neural network prediction model optimized by gravitation search algorithm and the RBF neural network prediction model; the simulation results show the prediction accuracy of RBF neural network optimized by improved gravitation search algorithm. Prediction model has improved greatly and thus proves the feasibility and effectiveness of the algorithm in the prediction field.