Methodological issues of improving loan accounting in digital banks

Abstract

This article investigates the methodological aspects of improving credit calculation in digital banks. The
study developed and tested a credit scoring model based on the gradient boosting algorithm. The results indicate
that alternative data — including clients’ digital activity and transaction patterns — is more effective in identifying
credit risk than traditional financial indicators. The model’s high accuracy enables optimization of credit policies,
expansion of financial inclusion, and early risk detection in digital banks. The study’s methodology and findings hold
both practical and theoretical significance, providing a basis for the implementation of innovative approaches in
banking.