IoT Network Administration by Intelligent Decision Support Based on Combined Neural Networks

At present, IoT networks have penetrated almost all spheres of life in modern society. They have a fairly wide arsenal of various network devices and also have a fairly developed and branched structure. However, the high dynamics of the behavior of IoT networks, coupled with the large volumes of information processed in them and the transmitted traffic, cause certain difficulties in solving the problems of administration of computer networks. It is becoming increasingly difficult for an IoT network administrator to identify and resolve abnormal situations in a timely manner. It is possible to solve the problem of effective administration of a large and complex IoT networks if we introduce a specialized intelligent decision support system for the administrator into the arsenal of network administration tools. The paper discusses a variant of the implementation of the analytical block for intelligent decision support of IoT network administrators, built on the basis of artificial neural networks. The paper outlines the structure of a combined neural network, focused on solving the problem of assessing the state of computer network elements. Three training methods are considered: stochastic gradient descent, the adaptive learning rate method, and the adaptive inertia method. The experimental results have shown a sufficiently high accuracy of the proposed solution, good adaptability, and the possibility of its application in a wide range of network configurations.