Scientific Journal of KubSAU

Polythematic online scientific journal
of Kuban State Agrarian University
ISSN 1990-4665
AGRIS logo UlrichsWeb logo DOAJ logo

Name

Lipin Konstantin Мikhailovich

Scholastic degree

Academic rank

Honorary rank

Organization, job position

Kuban State Technological University
   

Web site url

Email


Articles count: 2

Sort by: Date Title Views
177 kb

ANALYSIS OF METHODS AND MEANS OF EXTRACTING KNOWLEDGE AND ASSESSING THE QUALITY OF MODELS IN THE DSS FOR INDUSTRIAL-TECHNOLOGICAL AND CONSTRUCTION INDUSTRIES

abstract 1572003013 issue 157 pp. 170 – 181 31.03.2020 ru 180
The article considers the most used methods and means of knowledge extraction taking into account the quality assessment of models in decision support systems. In scientific and practical terms, the possibilities of joint effective use of expert systems, data mining (IAD) and machine logical inference (MLV), which provides deeper data processing, taking into account the significant differences between databases (DB) and knowledge bases (BZ). DB is a unit of information unrelated to each other information, while BZ – not only related to each other, but also with the concepts of the world, which makes it possible to solve complex multi-criteria problems in various subject areas. Currently, increasing attention is paid to non-network technologies that have the ability to simulate nonlinear processes, work with noisy data, as well as the ability to learn and self-study, extracting essential features from the incoming information. At the same time, the integration of neural network technologies and artificial intelligence models into a single hybrid system together with the methods of logical inference in the form of a hierarchical sequence of the "If-then" rules structure significantly improves the understanding of the studied process and the quality of presentation of the result. Nevertheless, these methods and means of knowledge extraction are insufficient if the fuzzy linguistic inference mechanism is not used. The basic characteristic of fuzzy sets is the membership function, which is a generalized characteristic of a normal set. To set this feature, we use three types of shapes – triangular, trapezoidal and Gaussian type and two main procedures – phasefication and de-phaseification which is considered by the example of the method of Mamdani. Along with the stated most promising direction in this area is the adaptive gain algorithm called AdaBoost, where the limitation of the gain due to the filtering is to apply the subsampling circuit which has the normal contour of batch training, reusable training data. This provides an opportunity to work with weak models, and in the conditions of hybridization causes efficiency increase, strengthens the classifiers united in the "Committee". Each next set of classifiers is built on objects incorrectly classified by previous sets. AdaBoost is sensitive to data noise and emissions and is less susceptible to retraining, which can significantly reduce the number of examples and obtain better output in the DSS
237 kb

THE ENGINEERING COMPLEXITY OF CONSTRUCTING THEORIES ON THE BASIS OF USE OF MEANS OF DISPLAY OF THE VARIOUS MODELS AND METHODS FOR DESIGNING DATA STRUCTURES

abstract 1582004003 issue 158 pp. 27 – 38 30.04.2020 ru 196
The article deals with the complexity of the construction of engineering theories in both scientific and practical direction based on the use of non-traditional approach to the problems of effective data monitoring, especially in the field of control systems of multicomponent representation of objects of system analysis. The study of the complexity of this issue involves a detailed consideration of the relationships of the elements of these objects on the basis of known methods, while the integration of heterogeneous knowledge obtained by such often independent methods becomes very time-consuming and poorly formalized. Currently, the processing of information and its subsequent presentation have changed significantly through the use of data mining (IAD), which includes not only the organization of the knowledge system in various missile defense, but also in the field of DSS. This, in turn, contributes to the effective formalization of fuzzy information and processing it in the form of fuzzy algorithms, which is an extension of the decision support system based on fuzzy logic – DSS NL. At the same time, it is necessary to emphasize the features of the proposed approach of the DSS NL, which is that it can be used in various missile defense systems, including for the effective analysis of statistical information of multicomponent representation of objects, which is used in determining statistical indicators to identify and assess existing and potential risks, adverse situations, as well as in the preparation of motivational grounds for managerial decision-making. For the purpose of more detailed establishment in real missile defense of the relations between objects it is offered to carry out by means of various degrees of dependence. For example, the types of graded connections are considered as fuzzy objective connections, and the use of expert systems and semantic links led to the construction of hypotheses analysis of situations and semantic relationship between them. A significant difference of the considered DSS NL is that each model is formed on the basis of a separate semantic network, and the system itself works with several models of Pro related or unrelated to each other. On the basis of the use of the concept of the relationship of proximity between concepts, belonging to the situation, its information part of the recommendations Are grouped according to the selected situation for their subsequent analysis and decision-making. On the basis of the principle of coordinating actions and construction of the function, taking into account the optimal time of the control action, the General algorithm of decision support for emergency production situations in the Pro low-rise construction, both in urban and rural areas
.