Mathematical Models of Time Series of Intensities of Network Traffic with Different Approaches to Their Creating

In this article the possibility of application of artificial neural networks and autoregressive analysis to creating prediction models the channel with packet data transfer describes. The problem is solved using example time row measurement intensities on port edge switch. Algorithm of building autoregressive models based on statistical analysis of time row measurements. Algorithm of neural network model on the basis of the definition of fractal dimension of time rows was shown. Built models compared to current values entered criteria for evaluating the accuracy of the prediction.

nonlinear dynamics, Takens theorem, neural networks, time rows, self-similar traffic of Gigabit Ethernet