Development of an adaptive management model for a production organization based on a neural network

In the conditions of economic instability of supplies, limited access to foreign tech-nologies and suppliers of raw materials and materials, as well as increasing pressure on the manufacturing sector of Russia, the issue of improving the management system of manufacturing organizations is especially acute. Operational logistics activities of a modern organization in-clude the management of financial activities, personnel, marketing, information technology, pro-curement logistics management, production logistics, sales, warehousing and return logistics. Traditional methods of managing the operational logistics activities of an organization are often not flexible enough to quickly respond to changing external and internal conditions, therefore, this study is aimed at eliminating the shortcomings in the management activities of the organiza-tion by developing an adaptive model for managing a manufacturing organization based on a neural network. As an intelligent tool, it is proposed to use a trainable neural network - GRU (Gated Recurrent Unit), capable of processing time series data and adapting to changes in con-trol parameters. The model being developed on the Streamlit platform using the Tensor-Flow/Keras framework is designed to forecast and automate management decision-making in op-erational activities and can be integrated into the practical activities of Russian manufacturing organizations. The article presents a comparison of GRU with alternative neural networks LSTM (Long short-term memory) and Transformer, which allows us to justify the choice of architecture in terms of efficiency and practical applicability.