T.V. Afanaseva, V.V. Voronina, A.A. Romanov

Periodic or continuous monitoring of the volume of traffic is one of the most important tasks of the operational support of computer networks. Dynamics of time series traffic depends on the time intervals observations and can be characterized by a high degree of uncertainty. For the administrator, the values of the traffic at a given moment of time have additional characteristic in the form of expert evaluation. The classical analysis of time series does not provide for modeling and forecasting of expert estimates. For these purposes, use fuzzy sets and models of time series in the form of a fuzzy time series. However, this model does not provide for the identification and modeling of fuzzy increment (fuzzy trends). To indicate fuzzy increment of fuzzy time series used introduced earlier by the authors of the concept of «fuzzy trend». The article provides the necessary definitions and properties of fuzzy trends, allowing the linguistic form to describe the behavior of the time series. Also contains the definition of the fuzzy process with fuzzy increments and model of time series. This model of time series alloy to model fuzzy trends of the computer network in the linguistic form. The experiments show the accuracy of short-term forecasting the volume of traffic and fuzzy trends.