UDC 004.89
NEURAL NETWORK METHOD FOR DATA RECOVERY IN INFORMATION SYSTEMS
A. D. Obukhov, Ph.D. (Tech.), associate professor of department «Automated decision support systems», TSTU, Tambov, Russia;
orcid.org/0000-0002-3450-5213, e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
M. N. Krasnyanskiy, Dr. Sc. (Tech.), professor, Rector of TSTU, Tambov, Russia;
orcid.org/0000-0002-8751-7445, e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
The problem of recovering missing or damaged data in information systems using neural network technologies is studied. The aim is to formalize and approbate the data recovery method, operating on the basis of the use of neural networks, which allows automating the process of determining the missing information of various types. The novelty of the proposed method lies in the automatic formation and training of an ensemble of neural networks with the selection of the optimal architecture and parameters for predicting the values of missing data based on the remaining information. The formalization of the neural network method of data recovery in the set theory notation is presented. The practical implementation of the method in the Python programming language is considered, experimental studies are carried out to assess its efficiency in comparison with the existing software solution (the Devol library). The developed method shows comparable accuracy, however, the time for solving the problem and the complexity of its software implementation are lower (by 53 %), and also allows obtaining a solution for all combinations of missing data through the use of an ensemble of neural networks. This confirms its effectiveness in solving problems of automating the processes of recovering lost data, which is highly relevant in the implementation of adaptive information systems and automatic control systems.
Key words: missing data recovery, neural networks, machine learning, ensemble of neural networks, set theory, adaptive information systems, neural network method.