Abstract: The subject of this research is to evaluate the effectiveness of the approximation of objects with discrete output based on fuzzy knowledge bases. The object of the research is the neuro-fuzzy model, which allows, based on the training of a fuzzy neural network, to form a system of fuzzy-production rules (a fuzzy knowledge base) for assessing the state of objects. The author examines in detail the type of fuzzy-production rules proposed by him, the algorithm of logical inference on the rules, describes the developed model of a fuzzy neural network. Particular attention is paid to the need to assess the approximating ability of the model in order to determine the feasibility and effectiveness of its practical use. This assessment was made by analyzing the following model characteristics:- convergence of the developed learning algorithm for fuzzy neural network;- satisfaction of its work with the principles of fuzzy approximation;- consistency of the logic inference algorithm on the rules of the model to the well-known algorithm for approximating objects with discrete output based on a fuzzy knowledge base. The estimation of the approximating ability of the neuro-fuzzy model was made, based on the results of which it was concluded that this model is an effective approximator of objects with a discrete output. In addition, in order to test the model, an assessment was made of the classifying ability of the fuzzy rules being formed. The accuracy of classification based on fuzzy rules turned out to be no lower than the accuracy of other known classification methods. The practical value of the application of such rules is the ability to build decision support systems for assessing the state of objects in various subject areas.

Abstract: The subject of the study is to evaluate the effectiveness of assigning tasks in the automated systems of electronic document management on the basis of fuzzy methods of rational choice of alternatives. The object of the research is the problem of a rational choice of employee for the task based on qualifications, performance, workload as well as difficulty of the assignment. The paper focuses on three job assignment strategies:  based on the method of maximin convolution, additive convolutions and fuzzy-production model. Particular attention is paid to the software developed on the basis of the proposed methods. The authors present an example of its operation, as well as the results of studies evaluating the effectiveness of allocation tasks for the performers. As the methods of research authors used methods of fuzzy rational choice of alternatives: maximin convolution, additive convolutions and fuzzy-production model. These methods are used to assign tasks the automated systems of electronic document management. The main conclusions of the study are:
-          method based on fuzzy logic inference has shown the best results, most closely consistent with the intuitive representation of an expert on rational choice assignments performers;
-          maximin convolution method is a pessimistic approach, that does not take into account the good side of the alternatives;
-          method of additive convolution is an optimistic approach in which low scores on the criteria have the same weight as compared to the high marks that affected its low accuracy.
A special contribution to the authors of the study is in developing an effective fuzzy-productions tasks assignment model as well as the implementation of the software system for performing the necessary calculations to evaluate its effectiveness. This determines the scientific novelty and practical value of the study.