Алгоритм минимального связующего дерева для оценки точек роста в системе стратегического планирования территорий

Abstract

The article considers modern topical approaches to strategic planning of economic development of territories using the concept of growth points and possibilities of graph model analysis. The aim of the study is to develop a method of spatial clustering of socio-economic units based on the analysis of graph models for the assessment of growth points in the system of strategic planning of territories. In addition, we present an approach to the construction of a generalized secondary graph of the municipal network of industrial enterprises based on the position of objects in the multidimensional feature space and the links between them using the UMAP algorithm, which allows us to perform a uniform approximation of the variety of variants of mapping objects in the multidimensional feature space with a correction for the distance to the nearest neighbor. The main hypothesis of the work is based on the graph complexity limitation by removing edges between vertices with low similarity degree and performing further clustering. We propose a minimum spanning tree (MST) selection method for determining growth points in a generalized secondary graph. The developed approaches can be successfully applied to improve the effectiveness of strategic planning of territories, but they require improvement. For this purpose, it is necessary to specify specific key indicators and indicators that will allow to form a secondary graph based on the UMAP algorithm to assess the interrelationships between the objects, as well as to eliminate noise and reduce overtraining of the model. Inclusion of new indicators that contribute useful information about the subject area into the model will allow scaling of the obtained solutions.