THE CONCEPT OF FORMING A DIGITAL TWIN STRUCTURE TO TAKE INTO ACCOUNT THE SIGNIFICANCE OF PURCHASED LITERATURE IN THE SYSTEM OF ACQUISITION OF THE BOOK COLLECTION OF THE UNIVERSITY LIBRARY

Abstract

The sphere of higher education, including library processes, nowadays functions in close interaction with information systems, among which information systems based on artificial intelligence technology seem to be especially effective. Implemented by the university educational process and scientific research are directly dependent on the qualitative and quantitative characteristics of the acquisition process of the book collection. In this regard, relevant is a possible way to use a digital twin, based on artificial intelligence technology, as a tool of intelligent management in the preparation of an order of literature for the university, by ranking the ordered publications on the indicators of informational relevance, based on the principles of information equivalence, and the unequal value of meaningful references. Research Objective. Discusses the structures of the proposed digital twin, examples of calculating the rankings of publications offered by publishers for purchase. Materials and methods. The authors analyze the principles of information equivalence, according to which all the significant references have the same informative value, and the principles of information unequalness, which consist in assigning informational weight (calculated on the basis of citation indices of authors of the refereed publication) to each reference made by the university staff. Results. Developed models of neural networks, which perform a digital twin and provide the possibility of determining the importance of the literature submitted for purchase, which is planned to be considered as the main in the process of implementation of the intelligent system of formation of the order of the literature, characterized by the optimum performance. We propose the use of a learning algorithm with a teacher, where the input is a training spectrum of data, which the neural network receives and then re¬cognizes dependencies and correctly responds to the incoming test data set. Once the neural network is fully trained, it is planned to disable the teacher, which will allow the neural network to work independently. Conclusion. The functionality of the developed abstract models of neural networks can be expanded to account the rating, qualitative (on the topics of publications), quantitative (by titles and types of publications, the cost of the order with delivery, etc.), as well as to determine the number of copies of the ordered publications.