Prediction for network traffic of radial basis function neural network model based on improved particle swarm optimization algorithm

Abstract

The neural network data are usually characterized by abruptness, nonlinearity and time variability, and thus, it is difficult to yield accuracy results of network traffic prediction based on a traditional radial basis function neural network that has the shortcomings of slow convergence and easily falling to local optimum. To improve the accuracy of network traffic prediction and optimize the method of parameter and structure setting for a neural network, a model of radial basis function neural network with improved particle swarm optimization algorithm is proposed by referring to the related theories of network traffic and phase space reconstruction. The improved particle swarm optimization algorithm can adjust the inertia weight and the learning factors, and make $$t$$ -distribution mutation of particles’ positions via global extremum to avoid local convergence and thereby improving its global searching capacity; with such an algorithm, the parameters of radial basis function neural network are optimized; then, in order to verify the algorithm’s effectiveness, the radial basis function neural network is trained to become an optimal prediction model, which is adopted for the prediction of two typical chaotic time series and the real network traffic. It is then compared with the traditional radial basis function neural network model and the radial basis function prediction model by improved particle swarm optimization; and the simulations result shows that the application of this model improves the accuracy of network traffic prediction, and demonstrates the algorithm’s feasibility and effectiveness for network traffic prediction.