Ukrainian Banks' Business Models Clustering: Application of Kohonen Neural Networks

Vladyslav Rashkovan and Dmytro Pokidin
Additional contact information
Vladyslav Rashkovan: SD Capital
Dmytro Pokidin: National Bank of Ukraine

Visnyk of the National Bank of Ukraine, 2016, issue 238, 13-38

Abstract: This paper clusters and identifies six distinct bank business models using Kohonen Self-Organising Maps. We show how these models transform over the crisis and conclude that some of them are more prone to default. We also analyze the risk profiles of the bank business models and differentiate between safest (valid) and riskiest ones. Specifically, six risk types (Profitability, Credit, Liquidity, Concentration, Related parties lending, and Money Laundering) are used to build risk maps of each business model. The method appears to be an efficient default prediction tool, since a back-testing exercise reveals that defaulted banks consistently find their place in a "risky" region of the map. Finally, we outline several potential fields of application of our model: development of an Early Warning System, Supervisory Review and Evaluation Process, mergers and acquisitions of banks.

Keywords: Neural networks; clustering; SOM; business model; banking (search for similar items in EconPapers)
JEL-codes: C45 G21 L10 (search for similar items in EconPapers)
Date: 2016
References: View references in EconPapers View complete reference list from CitEc
Citations: View citations in EconPapers (1)

Downloads: (external link)
https://journal.bank.gov.ua/en/article/2016/238/02 (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:ukb:journl:y:2016:i:238:p:13-38

DOI: 10.26531/vnbu2016.238.013

Access Statistics for this article

More articles in Visnyk of the National Bank of Ukraine from National Bank of Ukraine Contact information at EDIRC.
Bibliographic data for series maintained by Research Unit ().