UDC 004.855.5
APPROACH TO DETERMINATION OF SEARCH RANGE OPTIMAL VALUES OF PARAMETERS CLASSIFIER BASED ON THE FOREST OF DECISION TREES
L. A. Demidova, Dr. Sc. (Tech.), full professor, RSREU, Ryazan, Russia;
orcid.org/0000-0003-4516-3746, e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
M. S. Ivkina, post-graduate student, RSREU, Ryazan, Russia;
orcid.org/0000-0003-3677-1598, e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
The problem of determining the search ranges for the optimal values of classifier parameters based on the forest of decision trees (RF classifier, Random Forest classifier) with the aim of reducing the time spent on its development is considered. The aim of the work is to develop recommendations for determining the ranges of the search for values for such parameters of the RF classifier as the number of trees, the number of features by which the best splitting is found in a tree node, the depth of trees and the minimum number of objects at which the tree node is declared. The formation of recommendations is based on the results of experimental studies on the development of RF classifier models based on different data sets from machine learning data repositories. The results of experimental studies on the development of RF classifier models using training and test samples based on the analyzed data sets are given, graphical dependences on the assessment of the quality of classification on a test sample and the development time of the RF classifier on a training sample are obtained in general form and determining the search ranges for the optimal values of the RF classifier parameters.
Key words: classification, RF classifier, decision tree, random forest, optimal parameter value, search
range, classification quality indicator