ABSTRACTS OF ARTICLES OF THE JOURNAL "INFORMATION TECHNOLOGIES".
No. 9. Vol. 26. 2020

DOI: 10.17587/it.26.499-506

I. Ya. Lvovich, Professor, e-mail: office@vivt.ru, Ya. E. Lvovich, Professor, e-mail: office@vivt.ru, A. P. Preobrazhensky, Associate Professor, e-mail: app@vivt.ru, Voronezh Institute of High Technologies, Voronezh, Russian Federation, O. N. Choporov, Professor, e-mail: choporov_oleg@mail.ru, Voronezh State Technical University, Voronezh, Russian Federation

The Features of Machine Learning Methods

The paper provides a comprehensive analysis of the capabilities of approaches based on machine learning, describes their weaknesses and strengths. The types of machine learning are indicated. The subtask of supervised learning is considered on the example of the classification of objects. The parameters of the binary averaged perceptron are noted. The characteristics of an algorithm based on a binary Bayesian point machine are considered. The advantages and disadvantages of ensemble models are indicated. The features of a binary neural network are considered. The main characteristics of the logistic regression algorithm, advantages and disadvantages are noted. The tasks performed using this algorithm are considered, in which it is necessary to classify objects according to their characteristics into two classes: diagnostic problems, scoring models. The summary table compares the binary classification algorithms for accuracy, training time, linearity, adaptability, data volumes. The features of the Azure machine learning studio, in which the analyzed approaches are applied, are shown.
Keywords: machine learning, algorithm, data processing, neural networks, deep learning

To the contents