Search academic texts

Information × Registration Number 0313U005087, 0113U003106 , R & D reports Title "Research on the development of specialized software tools and efficient algorithms for optimization models on fuzzy logic and neural networks" popup.stage_title «Науково-дослідна робота з розробки спеціалізованих програмних засобів та ефективних алгоритмів оптимізації моделей на нечіткій логіці та нейронних мереж» Head Sventuh Andrew, Registration Date 20-02-2013 Organization Society with limited liability "Artificial Intelligence Laboratory" popup.description2 The object of study is the economic and mathematical model of diagnosing possible bankruptcy Ukrainian enterprises are built using tools of fuzzy logic, neural networks and discriminant analysis. The purpose of research is the development of specialized software tools and effective optimization algorithms mathematical economic models diagnosing possible bankruptcy Ukrainian enterprises built using tools of fuzzy logic, neural networks and discriminant analysis. Methods. When the GDR was applied gradient methods of optimization and stochastic models on fuzzy logic and neural networks, such as back propagation algorithm and the Extended Delta-Bar-Delta; highly specialized methods of setting these models, algorithms and Kohonen self-organizing maps parameters of Mamdani type fuzzy models, methods discriminant analysis and the center of gravity for the construction of discriminant models. Programmatic implementation of economic and mathematical models carried out in accordance with the recommendations of practitioners, standards and capabilities of modern software. To achieve the objectives of the study were delivered and settled scientific task improvement and practical implementation of optimization algorithms: рarameters of multilayer perceptron Rosenblatt; Kohonen self-organizing maps; neural Networks counter proliferation; parameters of mathematical economic models Mamdani type fuzzy logic; models of fuzzy logic knowledge base Sugeno; parameter discriminant models. All algorithms optimization mathematical economic models on fuzzy logic, neural networks and discriminant models brought to the software implementation. This shows the practical value of the study, as developed algorithms and software tools are universal and can be applied to adjust the economic and mathematical models of diagnosing possible bankruptcy Ukrainian enterprises within the project on GP/F44/027 "intelligent early warning system failures Ukrainian enterprises "and to optimize the parameters of other models on fuzzy logic, neural networks and discriminant models. Experimenting with developed software tools confirmed high efficiency of the existing optimization algorithms. Scope of the study results apply to the field of economic modeling as well as any other complex nonlinear processes and systems. Product Description popup.authors popup.nrat_date 2020-04-02 Close