Clustering Using the k-Means Method, Modeling the Stability of Russian Banking Companies Using the Dl-Model "Random Forest" in Modern Conditions

Abstract

The article discusses some theoretical aspects of modeling the sustainable development of the Russian banking system. The relevance of the study lies in the fact that in modern conditions, approaches to ensuring the sustainable development of the banking system using artificial intelligence are increasingly being used. The novelty lies in the fact that the authors have proposed approaches that allow us to identify existing patterns and formulate a forecast of an indicator of interest based on artificial intelligence. At the same time, the developed digital model involves its training on a generated dataset, including a data set that reflects the stability and dynamics of development of Russian banks. The work used such methods as monographic, analytical, k-means method, DL model "Random Forest" on the Colab service using the Python language and the libraries pandas, GridSearchCV, sklearn and others. The practical signifi cance of the study is that the resulting forecast can be used in practice regarding monitoring and forecasting the sustainable development of the banking system. The criterion for the forecast accuracy of the DL model is the average forecast error (MAE). The proposed DL model uses the best decision tree, which has optimal hyperparameter settings, for example, the depth of the tree is ten layers, the number of estimators (trees) in the ensemble is five.