Development of Financial Strategies Using Machine Learning for Right-Tail Value at Risk Estimation

Abstract

In this study, we propose a novel methodology for strategic financial planning based on the assessment of right-tail value-at-risk (VaR) for interest rates under volatile market conditions. The core of the approach lies in the integration of three model types: classical (GJR-GARCH), machine learning (XGBoost), and deep learning (long short-term memory, LSTM). Each component targets a distinct dimension of risk: analytical structure, nonlinear dependencies, and complex temporal patterns, respectively. The results showed that the LSTM model delivered the highest forecasting accuracy under structural instability, achieving the lowest residual volatility (σ̂ ≈ 0.0137) and a high level of explained variance (R² ≈ 0.69). An ensemble model (weighted 40% toward LSTM and 30% toward XGBoost) demonstrated superior reliability in risk estimation according to formal backtesting, along with the lowest average exceedance in VaR breaches. The practical value of the proposed approach is its ability to operate effectively under data scarcity and elevated volatility, enabling adaptive management of debt exposure. An elevated VaR level serves as a signal to restrict borrowing, while a low VaR opens opportunities for credit expansion. This model can be integrated into risk management systems of banking institutions and corporate finance departments as an effective tool for identifying, assessing, and mitigating right-tail risks during periods of economic instability.   Received: 2 July 2025 | Revised: 3 December 2025 | Accepted: 5 January 2026

Conflicts of Interest

The authors declare that they have no conflicts of interest to this work.

Data Availability Statement

The data and scripts that support the findings of this study are openly available in Figshare at https://doi.org/10.6084/ m9.figshare.30939428.

Author Contribution Statement

Vitaliy Makohon: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Writing – original draft, Writing – review & editing, Visualization, Supervision. Oleksandra Mandych: Conceptualization, Validation, Resources, Writing – review & editing, Visualization, Supervision. Tetiana Staverska: Methodology, Resources, Data curation, Writing – review & editing, Project administration. Oleksandr Horokh: Investigation, Resources, Writing – review & editing, Visualization. Denys Zabolotny: Conceptualization, Software, Validation, Investigation, Writing – original draft, Visualization.