CASPIAN JOURNAL
MANAGEMENT AND HIGH TECHNOLOGIES
Use of hybrid neural network models for mnogoagentny systems of classification in heterogeneous space of informative signs
 |
Read | Kurochkin Aleksandr G., Zhilin Valeriy V., Surzhikova Svetlana Ye., Filist Sergey A. Use of hybrid neural network models for mnogoagentny systems of classification in heterogeneous space of informative signs // Caspian journal : management and high technologies. — 2015. — №3. — pp. 85-95. |
Kurochkin Aleksandr G. - commercial director, LLC В«RPC" Innotech"В», 94 50 let Oktyabrya St., Kursk, 305040, Russian Federation, ak.kursk@gmail.com
Zhilin Valeriy V. - Ph.D. (Engineering), Associate Professor, Kursk State Agricultural Academy named after Professor I. I. Ivanov, 70 Karl Marks St., Kursk, 305004, Russian Federation, shatolg@mail.ru
Surzhikova Svetlana Ye. - post-graduate student, South-West State University, 19 Chelyuskintsi St., Kursk, 305004, Russian Federation, moi_lanchik@mail.ru
Filist Sergey A. - D.Sc. (Engineering), Professor, South-West State University, 19 Chelyuskintsi St., Kursk, 305004, Russian Federation, SFilist@gmail.com
To monitor the functional state of organs and systems of the person proposed to use the analysis of current-voltage characteristics of bio-active points, and then build multi-agent classifiers. To obtain the vector of informative features characterizing the state of current-voltage characteristics of bio-active points was used approximation of current-voltage characteristics of a polynomial of the seventh order. In order to diagnose the functional state of a biological object used multi-agent classifiers based on probabilistic neural networks and fuzzy neural networks. Classifiers contain three microslots, the first of which consists of modules of three-layer probabilistic neural networks, the second and the third module of two-layer fuzzy neural networks. The number of modules in these microsloths equal to the number of classes of the functional state of the systems studied. Classifiers allow us to determine the subjective probabilities of belonging of the input vector (array of vectors in the case of heterogeneous space of informative features) to be allocated to the classes of States of the studied objects.
Key words: вольтамперные характеристики биоматериалов, гетерогенное пространство информативных признаков, классификация состояний, многоагентные классификаторы, алгоритмы анализа, вероятностные нейронные сети, нечеткие нейронные сети, current-voltage characteristics