The use of self-organizing maps in rural areas clustering due to the level and conditions of regional development

Abstract

The article proposes a methodology for applying SOM (self-organising map) based on the average distance from sites to cluster centres and the Deductor analytical platform at the federal and regional level. The source data used is statistical information on the socio-economic development of rural areas by constituent entities of the Russian Federation for 2014-2019, provided on the Rosstat website under 9 groups of indicators: health, sports, tourism, trade, services, communications, investment, housing construction, housing conditions. The formation of the neural network resulted in the division of the subjects of the Russian Federation into 12 different clusters according to the indicators of socio-economic development of rural territories and the development of recommendations for the meaningful interpretation of the clusters. The analysis of these values shows a high differentiation of the rural areas in the constituent entities of the Russian Federation. In addition, the proposed methodology is piloted at the level of rural municipalities in a particular region to assess the degree of intra-regional disparities between rural areas. Authors revealed that within the Penza region there is more homogeneous rural development, as the neural network formed only 3 different clusters. The implementation of this methodology accounts for the possibility of previously unknown properties in the object and may lead to the formation of additional clusters reflecting new directions of rural development in the future.