Statistical Techniques vs. Machine Learning Models - A Comparative Analysis for Exchange Rate Forecasting in Fragile Five Countries

Aug 3, 2026

2013年美国联邦储备（Fed）结束扩张性货币政策，显著冲击“脆弱五国”（Fragile Five）。这些经济体高度依赖跨境资本流动，汇率传递效应导致通胀目标偏离，从而迫使货币当局与金融参与者需要更精确的汇率预测以减轻偏离并提升货币政策有效性。研究以脆弱五国为样本，比较传统统计方法与机器学习模型的预测表现。传统方法包括Naïve Drift、Theta、Holt指数平滑与ARIMA，机器学习包含RNN、LSTM、GRU与CNN。结果显示，机器学习在所有国家上均优于传统统计方法；传统方法方向性准确率约47%至60%，而RNN准确率约80%至90%。研究结论为相关宏观政策与交易决策提供参考。

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages18

Reading Time45 min

AccessFree to download

Introduction

Literature Review

Muhammed Raşid BAKIR, PhD Candidate (corresponding author) [E-mail: mrasidbakir@aksaray.edu.tr](E-mail: mrasidbakir@aksaray.edu.tr) Aksaray University, Turkey Professor İbrahim BAKIRTAŞ, PhD E-mail: [ibakirtas@aksaray.edu.tr](ibakirtas@aksaray.edu.tr) Aksaray University, Turkey Associate Professor Emre ÖLMEZ, PhDE-mail: [emre.olmez@bakircay.edu.tr](emre.olmez@bakircay.edu.tr) İzmir Bakırçay University, Turkey STATISTICAL TECHNIQUES VS. MACHINE LEARNING MODELS: A COMPARATIVE ANALYSIS FOR EXCHANGE RATE FORECASTING IN FRAGILE FIVE COUNTRIES Abstract. In 2013, the Federal Reserve (Fed) announced the end of its expansionary monetary policy, which had a significant impact on certain countries. These countries, colloquially referred to as the "fragile five", were heavily dependent on financial capital flows, which led to deviations from inflation targets due to the exchange rate pass-through effect. Consequently, monetary authorities and other financial actors need accurate exchange rate forecasts to mitigate these deviationsand improve the effectiveness of monetary policy. This study aims to forecast the exchange rates of the fragile five countries using both traditional statistical methods and machine learning techniques. The traditional statistical methods used in this study include Naïve Drift, Theta, Holt's Exponential Smoothing and ARIMA models, while the machine learning methods include RNN, LSTM, GRU and CNN architectures. The results show that machine learning methods outperform traditional statistical methods in terms of prediction accuracy for all countries. While statistical methods show a directional accuracy rate between 47% and 60%, RNN, one of the machine learning models, shows an accuracy rate between 80% and 90%. Overall, these results suggest that machine learning methods can provide more accurate exchange rate forecasts for the fragile five countries than traditional statistical methods. These findings may be valuable for monetary authorities and financial actors seeking to improve the effectiveness of monetary policy in these countries. Keywords: machine learning, forecasting methods, emerging market, monetary policies, exchange rate JEL Classification: E47, G17, F31, F37, E52 295 DOI: 10.24818/18423264/[57.3.23.18](57.3.23.18) 1. Introduction The year 2013 marked a significant period in global financial markets, particularly for developing countries. The Federal Reserve's decision to end its expansionary monetary policy led to significant capital outflows from developing countries, which increased exchange rate volatility (Nechio, 2014) . A report by Morgan Stanley further indicated that Brazil, India, Indonesia, South Africa, and Turkey were the most vulnerable to the Fed's policy adjustments due to their high inflation rates and current account deficits. These countries were therefore referred to as the 'fragile five' and financial investors were advised to trade the US dollar position against their currencies. The Fed's policy shift and the Morgan Stanley report underscored the importance of reliable exchange rate forecasts for monetary authorities and financial investors. The fragile five countries rely heavily on financial capital flows, which make their currencies volatile. This volatility creates many trading opportunities for financial investors. However, it also poses a challenge for monetary authorities, who have to take into account foreign exchange liabilities, forward options, and swap transactions that can affect macroeconomic variables. In addition, the pass-through effects of the exchange rate on imported intermediate and final goods may deviate from the target inflation rate and jeopardise the legitimacy of the monetary authority (Fendoğlu, 2020; Takhtamanova, 2010) . As a result, monetary authorities and other financial market players are forced to develop predictive models to reduce the exchange rate risk. As the use of machine learning (ML) methods increases in

FAQs

研究为什么聚焦“脆弱五国”的汇率预测？

脆弱五国高度依赖资本流动，汇率波动及汇率传递效应可能使通胀偏离目标，进而影响货币政策效果，因此需要更可靠的预测模型。

传统统计方法与机器学习方法在本研究中分别有哪些模型？

传统统计方法包括Naïve Drift、Theta、Holt指数平滑与ARIMA；机器学习方法包括RNN、LSTM、GRU以及CNN架构。

结果表明哪类方法在预测准确度上更优？

机器学习方法在所有国家的预测准确度上优于传统统计方法；传统方法方向性准确率约47%至60%，RNN准确率约80%至90%。

Aug 3, 2026

Upload by Lucas Martin