: ..
: -
:
:
: 2020
: ..
:
: 88
: .. - // . 88. .: , 2020. .99-123. DOI: https://doi.org/10.25728/ubs.2020.88.5
:
: , , , , ,
(.): hierarchical system of fuzzy inference, normalization of input variables, clustering, Hausdorff metric, object classifier, mountain clustering algorithm
: , . . , . . . , , . . . , , . , , . .
(.): Provisions for the improvement of computing systems based on the use of hierarchical fuzzy inference systems associated with preliminary data preparation and subsequent interpretation of the results are proposed. Data preparation is based on performing operations to normalize the input parameter values to a scale with the same range of values, provided that these values have a positive correlation with the values of the output variable. The implementation of the provisions is based on the use of piecewise functions. The use of provisions makes it possible to simplify the formation of production rules in the knowledge base of the fuzzy inference system. The study of the behavior of changes in estimates depending on the number of levels of hierarchies in the computing system made it possible to identify a property associated with the grouping of output estimates in the vicinity of the values of the terms that describe the output variable. This made it possible to refer the hierarchical system of fuzzy inference to the classifier of the analyzed objects by their state, taking into account the criteria of the assessment. Belonging to a certain class is determined by the value of the integral final grade. For the identify groups of objects with similar properties within classes, a set of operations is proposed based on the sequential use of the mountain clustering algorithm, the Euclidean metric and the Hausdorff metric. The use of operations makes it possible to single out typical representatives of the studied classes, and then determine the objects that are closest to them in terms of parameters, taking into account the established restrictions on deviations. The experiment carried out confirmed the efficiency of the proposed provisions.
PDF
: 1793, : 400, : 11.