Packet switching networks traffic prediction based on radial basis function neural network

Abstract

New multimedia applications require Quality of Service support, which is still not successfully implemented in current packet-switched networks implementations. This paper presents a concept of neural network predictor, suitable for prediction of short-term values of traffic volume generated by end user. The architecture is Radial Basis Function neural network, optimized with respect to a number of neurons. Testing mode of the neural network is very fast, what enables application of this tool in nodes of telecommunication network. This would help to warn a network management system on early symptoms of congestion expected in the near future and avoid the network overload.