TRADITIONAL MODELS OF MACHINE LEARNING IN THE FIELD OF THE INTERNET OF THINGS

Abstract

The article investigates traditional models of machine learning in the field of the Internet of Things. The directions
of distribution of intelligent systems are determined and the prospects of development are analyzed. Describes the objects of
everyday life that belong to the group of the Internet of Things. It is emphasized that the huge scale of IoT networks brings
new tasks, such as managing these devices, huge amounts of data, storage, communication, computing, security and privacy.
The cornerstone of the commercialization of IoT technologies is the guarantee of security and confidentiality, as well as
consumer satisfaction. The main obstacles to improving the safety of smart devices include market competition and technical
constraints. It is emphasized that traditional machine learning models are aimed at characterizing and determining the level
of harmful effects of IoT devices, training and testing of the neural network to classify IoT devices based on network traffic to
provide classification of IP addresses, close to real time and evaluate algorithms. The classification of traditional machine
learning algorithms is carried out: learning algorithms with teacher, without teacher and reinforcement. Each of them is
described. The principles of deep learning and reinforced learning, as well as their combination are defined. It is emphasized
that in the real world, when data from different sources have different formatting and representation, the constant principle
of the primary data set does not work and machine learning requires pre-processing and purification of data before placing
them in a particular model.