Tareq Y. Computational methods for teletfaffic models identification.

The dissertation is devoted to reducing the complexity of modelling time series to predict the teletraffic characteristics using expert system of production type. Teletraffic modelling plays a significant role in designing and managing computer networks. However, the properties of non-stationarity, self-similarity, and nonlinearity of its characteristics constraints the application of effective methods of stochastic prediction, designed for stationary or special cases of non-stationary time series. In this connection automation of constructing prediction for non-stationary time series is difficult to implement, and developing methods to identify models of non-stationary time series of teletraffic characteristics is an urgent task. To identify the models of teletraffic characteristics the software prototype of automated identification system based on a production expert system was developed.