Forecasting Railway Network Data Traffic: A Model and a Neural Network Solution Algorithm

Forecasting network data traffic is an important part of the function of planning and managing information systems. However, the contents of network data are so stochastic and complex that it is very difficult to establish stable functions to describe the mapping relationship between data flows and associated causal influences. In this paper, a multi-layer feed forward neural networks (NN) model is put forward to identify such relationship and the corresponding learning rule of NN, back-propagation (BP) algorithm, is given. In addition necessary estimation and validation processes are designed to ensure the successful implementation of the model proposed. The paper elucidates the application of NN model around the case of forecasting China railway Transportation Management Information Systems (TMIS) network traffic. The predictive results obtained demonstrate that the NN model and the solution algorithm are applicable for information planning on the TMIS network.