This paper describes the development of a recommender system for the selection and configuration of deep neural network models. Her key area of work is evaluating the effectiveness of a neural network for a particular type of problem. The recommender system designed within the framework of the repository allows for precise search, selection and flexible configuration of neural networks depending on specific project-oriented tasks.

Keywords: recommender system, machine learning, deep learning, neural network model, expert system, neural network, repository, spatial data, geographic information system, configuration, architecture

The article describes the development of a repository of neural network models designed to solve design problems in the field of spatial data analysis. The key options for using the system to provide support for making managerial decisions in the field of sustainable development are presented. The results of designing the ontological model of the repository are presented. The storage subsystem for neural network models is implemented in the form of a metalanguage. Models are converted into representations that are compatible with modern software systems for deep machine learning. The graphical web interfaces of the repository of deep neural network models and the visualization subsystem of deep machine learning models in the form of graph diagrams, as well as interfaces for obtaining structured information about specific neural network models are given. Within the framework of the repository, there is a subsystem for delimiting access rights for administrators and users of the system. The issues of updating the repository of deep neural network models in the process of solving practice-oriented problems in the field of ensuring the conditions for sustainable development of Russian regions and developing a recommender system for selecting and configuring neural network models stored in the repository are discussed.

Keywords: repository, neural network, neural network model, deep learning, spatial data, geographic information system, interface, data exchange, object detection, application programming interfaces

This paper presents the principle of operation of the algorithm for translating a graphical representation of neural network models into a program code within the framework of a neural network repository. Based on the obtained data structure, it is possible to carry out sequential generation of program code in a general-purpose programming language.

Keywords: neural network model, neural network, repository, graph model, programming, translation, spatial data, algorithm, topology, architecture