This article is cited in 3 papers

Logical methods of correct data classification

E. V. Djukova, G. O. Masliakov, A. P. Djukova

Federal Research Center “Computer Science and Control” of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119133, Russian Federation

Abstract: The work is devoted to issues of the application of discrete apparatus (logical methods of integer data analysis) for the supervised classification problem. Three main directions of logical classification are considered: Correct Voting Procedures (CVP), Logical Analysis of Data (LAD), and Formal Concept Analysis (FCA). Using the terminology of the CVP direction, the basic concepts applied in LAD and FCA are presented. The general scheme of the logical classifier is described, according to which each logical classifier at the training stage sets some partial order on a special set of fragments of precedents descriptions and searches for the maximum elements of this set relative to the given order. Such studies are important for creating a general theory of correct classification according to precedents based on the use of a discrete apparatus.

Keywords: supervised classification, logical classifier, Correct Voting Procedures, Logical Analysis of Data, Formal Concept Analysis, irredundant representative elementary classifier, strong pattern, JSM-hypothesis, partial order.

Received: 11.01.2023

DOI: 10.14357/19922264230309