ABSTRACTS OF ARTICLES OF THE JOURNAL "INFORMATION TECHNOLOGIES".
No. 10. Vol. 24. 2018

DOI: 10.17587/it.24.665-670

N. A. Ignatiev, D. Sc., Professor, e-mail: n_ignatev@rambler.ru, National University of Uzbekistan named after M. Ulugbek, Tashkent

Selecting an Object's Own Space Using Nonlinear Transformations of Features

The description of objects in their own space from latent features is considered. For the formation of the space, the rules of a hierarchical agglomeration grouping of initial features are used. The rules provide the property of invariance to the scales of measurement for non-linear mapping of groups of raw polytypical (nominal and quantitative) features onto the numerical axis. The need to apply non-linear displays of initial features to latent ones according to the rules of the hierarchical agglomeration group is explained as follows:
—from with some (initially unknown) dimension of the source features space, the relations of proximity between the objects of classes become blurred;
— the structure of relations between objects depends on the scale of measurements of quantitative features and on the adjacency measures used.
As a separate problem, the choice of informative latent features of the object is considered. The description of an admissible object within a space of informative latent features is necessary for finding an individual measure of similarity (difference) with other objects. This measure reflects the relationship between the objects and serves as a means for decision-making. It is practically impossible to obtain a comparable effect from the methods of selecting sets of features that are informative on the whole training sample. The technique of selection of an informative set of features of an object by three criteria is described. Measure of similarity in this set allows relate the object to the number of standards or to identify it as noisy. The prospects of discovering and interpreting hidden regularities in databases from unified methodological positions are revealed. For example, in the field of medical computer diagnostics, one can choose an individual course of treatment for a patient.

P. 665–670

To the contents