Time-series XAI for FX rate or inflation-linked risk forecasting: Transformer forecasting

¹ Tashkent State Agrarian University
² Termez University of Economics and Service
³ Urgench State University named after Abu Rayhon Beruni
⁴ Tashkent University of Architecture and Civil Engineering
⁵ Termez State University

^* Corresponding author’s: This email address is being protected from spambots. You need JavaScript enabled to view it.

Published online: 13 May 2026

Abstract

Many emerging market participants are facing challenges of not having reliable early-warning signals to maintain their portfolio stability upon exchange rate shocks and this vulnerability is becoming persistent over time. The study aims to integrate transformer forecasting and vector autoregression with the objective of improving risk prediction to support inflation-linked exposure management. The purpose of this research is to determine the predictive capacity of macroeconomic and financial resources linked to the foreign exchange market which are based on time series evidence. A multivariate time series framework was used to analyze structural dynamics contributing to the forecasting performance of exchange rate volatility for inflation risk, currency depreciation, liquidity stress and survival probability estimation. Data was collected among monthly observations of exchange rate and inflation indicators (CPI) and analyzed using vector autoregression and parametric survival models. The findings indicate that macroeconomic instability, inflation persistence, and external shock transmission channels have significant impacts on forecasting effectiveness with the explainable support of attention mechanisms. The results further indicate that structural factors like interest rate-setting, cross-border capital exposure and inflation expectation shifts affect financial market participants in enhancing risk anticipation capacity, which eventually improves portfolio resilience. This generate an implication that the experiences of a supportive predictive modeling environment lead investors to affectively feel confident to their decision processes, hence, strengthen their capacity to manage financial uncertainty for their assets, liabilities and long-term planning.