STUDY OF THE POSSIBILITY OF USING MACHINE LEARNING METHODS TO PREDICT THE DEVELOPMENT OF BRANCHES OF SCIENCE

Abstract

Статья посвящена определению принципа использования библиометрических показателей при создании подсистемы прогнозирования публикационной активности с использованием интеллектуальных методов. Показано, что наиболее оптимальным является метод машинного обучения DecisionTree. The article is devoted to determining the principle of using bibliometric indicators when creating a subsystem for predicting publication activity using intelligent methods. Key bibliometric indicators and examples of calculations for three areas within the technical scientific direction are presented: neural networks, genetic algorithms and data science. Machine learning methods for forecasting under uncertainty are analyzed. The choice of tools for the development of a mathematical model of the decision-making process in the field of scientific research has been made. It is shown that the DecisionTree machine learning method is the most optimal. The methodological apparatus is described, a training sample is created, and a predictive mathematical model is proposed that allows one to speak with sufficient accuracy about the number of future publications within each selected scientific group.