MULTIDIMENSIONAL DATA ANALYSIS IN THE TASK OF PREDICTING THE ENTRY OF CREDIT INSTITUTIONS INTO THE RISK ZONE

The study of economic processes is based on the study of a large number of parameters. In this connection, in order to analyze the phenomena under study and solve prognostic problems, there is a need to use methods of multidimensional data analysis. The article examines the problem of identifying suspicious, from the point of view of financial solvency, credit institutions operating in the Russian market. This study is aimed at developing a methodology for multidimensional data analysis to identify suspicious credit institutions and predict the revocation of their licenses. To solve this problem, it is proposed to use hierarchical and iterative methods of cluster analysis, as well as the principal component method. Based on these methods, a methodology for forming a risk zone has been developed that makes it possible to predict the revocation of licenses from credit institutions. To determine the number of clusters, the Ward clustering method was used, as well as the elbow method, the silhouette method, and the Davis-Bouldin method. The combined use of cluster analysis methods and the principal component method made it possible to demonstrate the robustness of the proposed methodology. Data from Bank Reporting Form No. 101 were used in this study.

1. Anderson T. Vvedenie v mnogomernyj statisticheskij analiz. [Introduction to multivariate statistical analysis]. Moscow, State Publishing House of Physical and Mathematical Literature Publ., 1963. 500 p.

2. Durand B., Odell P. Klasternyj analiz. [Cluster analysis]. Moscow, Statistika Publ., 1977. 128 p.

3. Dr. Tirthajyoti Sarkar. Clustering and dimensionality reduction techniques combined.

4. Available at: https: //github.com/tirthajyoti/Machine- Learning-with-Python/blob/master/Clustering-Dimensio¬nality-Reduction/Clustering_with_dim_reduction.ipynb (accessed 07.07.2023).

5. Ajvazyan S.A., Buhshtaber V.M., Enyukov I.S. Prikladnaya statistika. Klassifikaciya i snizhenie razmernosti [Applied statistics. Classification and dimension reduction]. Moscow: Finansy i statistika Publ., 1989. 607 p. ISBN 5-279-00054-X.

6. Akishina E.P., Ivanov V.V., Kryanev A.V., Prikazchikova A.S. Cravnitel'nyj analiz metodov derev'ev reshenij i nejronnyh setej v zadache klassifikacii kreditnyh organizacij [Comparative analysis of decision tree and neural network methods in the problem of classification of credit institutions]. Vestnik NIYaU MIFI, 2022. Vol. 11. No. 6. Pp. 442–449. https://doi.org/10.26583/ vestnik.2022.12 (in Russian)

7. Akishina E.P., Ivanov V.V., Prikazchikova A.S. Primenenie nejronnyh setej i metoda glavnyh komponent dlya identifikacii kreditnyh organizacij, potencial'no vovlechennyh v process po legalizacii prestupnyh dohodov [Application of neural networks and the principal component method for identifying credit institutions potentially involved in the process of money laundering]. Izvestiya Issyk-Kul'skogo foruma buhgalterov i auditorov stran Central'noj Azii, 2022. No. 2(37). Pp. 294–296 (in Russian).

8. Murtagh F. Ward's Hierarchical Agglomerative Clustering Method: Which Algorithms Implement Ward's Criterion. Journal of Classification, 2014. No. 31. Pp. 274–295.

9. Ayvazyan S.A. Mnogomernyj statisticheskij analiz v social'no-ekonomicheskih issledovaniyah [Multivariate statistical analysis in socio-economic research]. Ekonomika i matematicheskie metody, 1977. No. 13 (5). Pp. 968–985 (in Russian).

10. StatSoft Inc., 2011. STATISTICA (data analysis software system), version 10. Available at: www.statsoft.com (accessed 07.07.2023).

11. Baimuratov I.R. Metody avtomatizacii mashinnogo obucheniya [Methods for automating machine learning[. St. Petersburg, Universitet ITMO Publ., 2020. 40 p.

12. Ajvazyan S.L., Enyukov I.S., Meshalkin L.D. Prikladnaya statistika. Issledovanie zavisimostej [Applied statistics. Dependency research]. Moscow, Izdatel'stvo «Finansy i statistika» Publ., 1985. 488 p.