Оценка количества малых и средних предприятий в границах городов и агломераций: применение методов машинного обучения к данным России

Abstract

Introduction: the development of agglomerations, which, as research shows, are becoming productivity growth poles, is one of the priorities in the spatial development policy in Russia. Despite numerous discussions, the term “urban agglomeration” is currently absent in federal legislation, but it is widely used in the regulatory framework at other levels of government. This leads to a lack of coordinated, accurate, and up-to-date information on the size and structure of agglomerations, thus bringing forward the need to formulate a well-based approach to identify their economic and geographical boundaries. Objectives: developing an approach to identify the boundaries of agglomerations in Russia, considering the spatial distribution of economic activity. Methods: machine learning, statistical analysis. The use of these methods makes it possible to identify the boundaries of agglomeration with high accuracy using open data on the height of the urban area. Results: the article presents the results of spatial clustering algorithm (DBSCAN) testing on residential development data in Russia; a list of “sustainable municipalities clusters” defining the agglomerations boundaries has been identified and estimates have been made for the number and revenue of firms and SMEs within the largest cities and agglomerations boundaries defined both analytically and formally in regulatory documents. Using the example of individual regions and average data for Russia, it has been shown that the number of firms (total and those registered as SME) within agglomerations can be underestimated by more than one-third and total revenues – by almost 60 %. Conclusions: considering sustainable municipalities clusters can improve the accuracy of agglomeration effects and business activity estimates and contribute to the formation of more accurate government support measures.