CASPIAN JOURNAL
MANAGEMENT AND HIGH TECHNOLOGIES
Virtual flows in hybrid decision modules of classification of complex-structured data
 |
Read | Kiselyov A. V., Savinov D. Yu., Filist S. A., Shatalova O. V., Zhilin V. V. Virtual flows in hybrid decision modules of classification of complex-structured data // Caspian journal : management and high technologies. — 2018. — №2. — pp. 137-149. |
Kiselyov A. V. - Teacher of Information Systems and Technologies Department, Southwest State University, building B, 19 Chelyuskintsev St., Kursk, 305004, Russian Federation, Kiselevalexey1990@gmail.com
Savinov D. Yu. - post-graduate student, Southwest State University, building B, 19 Chelyuskintsev St., Kursk, 305004, Russian Federation, marina-savinova-93@mail.ru
Filist S. A. - Doct. Sci. (Engineering), Professor, Southwest State University, building B, 19 Chelyuskintsev St., Kursk, 305004, Russian Federation, SFilist@gmail.com
Shatalova O. V. - Cand. Sci. (Engineering), Associate Professor, Southwest State University, building B, 19 Chelyuskintsev St., Kursk, 305004, Russian Federation, shatРѕlg@mail.ru
Zhilin V. V. - Cand. Sci. (Engineering), Associate Professor, Kursk Institute of Cooperation, branch of Belgorod University of Consumer Cooperation, Economics and Law, 116 Radishchev St., Kursk, 305004, Russian Federation, vvzhilin61@gmail.com
For classification of difficult structured data hybrid decisive modules with virtual streams are offered. Virtual streams reflect the hidden system communications between observed parameters of process or system. At the same time the vector of informative signs consists of two subvectors. First of which corresponds to real streams, and the second - to virtual streams. The generalized recurrent block diagram is developed for formation of a virtual stream. The scheme allows to form the hybrid vector of informative signs consisting of two subvectors. One of which consists of initial informative signs, and the second - of the informative signs, received on the basis of modeling of system communications between informative signs of the first subvector. For formation of the second subvector again created latent variables are used as system communications between initial space of informative signs. That allows to realize recurrent process of informative signs space formation. The offered method of nonlinear models formation for virtual streams is based on MGUA-modeling. For receiving models of real streams influence in the proposed method are used the virtual streams neural networks on nonlinear adalina. The method allows to form a subvector of latent variables of unlimited dimension. On the basis of the offered method the structure of hybrid decisive system with virtual streams, intended for classification of difficult structured data, is created. The structure allows to consider latent informative signs (virtual streams), defined on the basis of statistical and expert researches of communications between initial informative signs. In turn it gives the chance to aggregate accurate and indistinct decisive rules, providing the required quality of the decision making in the situations of diverse and badly formalizable structure of classes. Simulation showed the effectiveness of the proposed method.
Key words: гибридный, решающая система, латентная переменная, МГУА-модель, нейронная сеть, нечеткая логика принятия решений, агрегаторы нечетких решающих правил, hybrid decision module, latent variable, GMDH model, neural network, fuzzy logic of decision-making, agg