At present, forecasting the state of Telecom networks is the most important task of network administration. Characteristic features of the network are the high dynamics of processes that occur in it, as well as the structural complexity of connections between nodes. They determine the need to use tools that are highly adaptable and resistant to external influences to assess the state of the network. Neural networks can be used as such tools. The neural network performs logical inference based on the fuzzy logic apparatus, otherwise there is no new knowledge and uses it in further work. Such systems are successfully used in many fields (machine learning, facial recognition, medicine, signal processing, military, and others), they allow you to get the desired results without human involvement and with little computational cost.