



























Abstract:Accurate forecasting of exchange rates remains a persistent challenge, particularly for emerging economies such as Brazil, Russia, India, and China (BRIC). These series exhibit long memory and nonlinearity that conventional time series models struggle to capture. Exchange rate dynamics are further influenced by several key drivers, including global economic policy uncertainty, US equity market volatility, US monetary policy uncertainty, oil price growth rates, and short-term interest rates. These empirical complexities underscore the need for a flexible framework that can jointly accommodate long memory, nonlinearity, and the influence of external drivers. We propose a Neural AutoRegressive Fractionally Integrated Moving Average (NARFIMA) model that combines the long memory structure of ARFIMA with the nonlinear learning capability of neural networks while incorporating exogenous variables. We establish asymptotic stationarity of NARFIMA and quantify forecast uncertainty using conformal prediction intervals. Empirical results show that NARFIMA consistently outperforms benchmark methods in forecasting BRIC exchange rates.
| Subjects: | Econometrics (econ.EM); Machine Learning (cs.LG); Applications (stat.AP); Machine Learning (stat.ML) |
| Cite as: | arXiv:2509.06697 [econ.EM] |
| (or arXiv:2509.06697v2 [econ.EM] for this version) | |
| https://doi.org/10.48550/arXiv.2509.06697 arXiv-issued DOI via DataCite |
From: Tanujit Chakraborty [view email]
[v1]
Mon, 8 Sep 2025 13:49:48 UTC (675 KB)
[v2]
Tue, 12 May 2026 09:50:55 UTC (661 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理,仅供阅读参考。 原文来自 — 版权归原作者所有。