Open AccessStudying the capabilities of the analytical system based on the machine learning method
Author(s) -
S. V. Palmov
Publication year - 2020
Publication title -
radiopromyšlennostʹ
Language(s) - English
Resource type - Journals
eISSN - 2541-870X
pISSN - 2413-9599
DOI - 10.21778/2413-9599-2020-30-3-112-126
Subject(s) - computer science , robustness (evolution) , machine learning , reliability (semiconductor) , decision tree , quality (philosophy) , artificial intelligence , variance (accounting) , data mining , biochemistry , chemistry , power (physics) , physics , philosophy , accounting , epistemology , quantum mechanics , business , gene
Data analysis carried out by machine learning tools has covered almost all areas of human activity. This is due to a large amount of data that needs to be processed in order, for example, to predict the occurrence of specific events (an emergency, a customer contacting the organization’s technical support, a natural disaster, etc.) or to formulate recommendations regarding interaction with a certain group of people (personalized offers for the customer, a person’s reaction to advertising, etc.). The paper deals with the possibilities of the Multitool analytical system, created based on the machine learning method «decision tree», in terms of building predictive models that are suitable for solving data analysis problems in practical use. For this purpose, a series of ten experiments was conducted, in which the results generated by the system were evaluated in terms of their reliability and robustness using five criteria: arithmetic mean, standard deviation, variance, probability, and F-measure. As a result, it was found that Multitool, despite its limited functionality, allows creating predictive models of sufficient quality and suitable for practical use.