Assessment of factors of regional economic stability using the XGBoost algorithm

Abstract

In modern conditions, the regions of Russia acquire a special role and are evaluated as independent economic entities. In a pandemic, the key factor for successful development at the meso-level is not so much the growth of well-being, the quality of life, but the preservation of the stability of economic systems, their ability to withstand external influences. Thus, in our study, we equate the economic stability and economic security of the regions. To assess economic stability, it is advisable, in our opinion, to use a group of indicators characterising resource provision, investment climate and the efficiency of functioning of regions. We assigned the rank of economic security to all regions, on the basis of which the regions were divided into two classes – economically safe and economically unsafe. The ensemble machine learning algorithm XGBoost was chosen as a method for factor analysis of economic security. The constructed classification model was interpreted by us using the Shap algorithm, which assumes the analysis of Shapley values for each economic determinant. The applied algorithms allowed us to identify significant factors that determine stable regions. These factors include investment risk, human development index and the ratio of the balanced financial result to the gross regional product.