Deterministic convergence of an online gradient method for neural networks
Journal of Computational and Applied Mathematics - Selected papers of the international symposium on applied mathematics, August 2000, Dalian, China
Analysis of input-output clustering for determining centers of RBFN
IEEE Transactions on Neural Networks
On overfitting, generalization, and randomly expanded training sets
IEEE Transactions on Neural Networks
| Hi-index | 0.00 |
Network traffic prediction is an important research aspect of network behavior. Conventionally, ARMA time sequence model is usually adopted in network traffic prediction. However, the parameters used in normal time sequence models are difficult to be estimated and the nonstationary time sequence problem can not be processed using ARMA time sequence model. The neural network techniques may memory large quantity of characteristics of data set by learning previous data, and is suitable for solving these problems with large complexity. IP6 network traffic prediction is just the problem with nonlinear feature and can be solved using appropriate neural network model. In this paper, according to the daily cycle characteristic of IPv6 network traffic, a novel transfer function is designed, which has lots of advantages such as fast convergence and high precision. Based on the new transfer function, an improved BP neural network model is produced, and a IPv6 network traffic prediction system is implemented. Using this new BP neural network model to process the actual data, the results present that our model has a faster learning ability and has a higher precision compared with previous BP neural network model. Therefore, this BP neural network model can be used for normal traffic prediction in current IPv6 network.