A Comparative Analysis of Machine Learning Algorithms for USD/EUR Foreign Exchange Rate Forecasting

• Author(s): Obi-Obuoha Abiamamela ; Mubaraq Onipede ; Owa Olorunsola Jason
• Paper ID: 1706704
• Page: 443-455
• Published Date: 20-12-2024
• Published In: Iconic Research And Engineering Journals
• Publisher: IRE Journals
• e-ISSN: 2456-8880
• Volume/Issue: Volume 8 Issue 6 December-2024

Abstract

The Forex market is a beast—complicated, ever-changing, and influenced by an overwhelming number of factors. Yet, in the chaos lies opportunity. With the advancement of machine learning, are we finally capable of taming this beast? This study dives deep into the comparison of four machine learning models—Artificial Neural Network (ANN), Autoregressive Integrated Moving Average (ARIMA), Random Forest, and Support Vector Machine (SVM)—to predict the hourly price movements of the EUR/USD currency pair. The dataset, sourced from MetaTrader 5, spans over three years and was engineered to include domain-specific indicators like RSI, Bollinger Bands, and moving averages. ARIMA performed well at spotting linear trends but couldn’t handle the market's unpredictable, non-linear flips. ANN stepped up, proving its worth in capturing these chaotic patterns, especially during sudden trend changes. Random Forest, though brilliant in training, suffered a severe case of overfitting when tested. As for the SVM models, the Linear and RBF kernels showed balance, while the Polynomial and Sigmoid kernels stumbled in tackling the intricate dance of Forex data. Each model had its highs and lows: ARIMA’s precision with linear trends, ANN’s knack for non-linear shifts, Random Forest’s training prowess but testing pitfalls, and the SVM’s moderate but stable performance. The takeaway? ANN shines brightest in volatile markets, ARIMA is the go-to for linear trend detection, and a hybrid approach combining their strengths might just be the silver bullet. In conclusion, this study underscores the importance of picking the right tool for the job in the Forex market. Careful parameter tuning, feature engineering, and even hybrid models could hold the key to consistently outsmarting the market.

Keywords

Autoregressive Integrated Moving Average, Support Vector Machine, Random Forest, Artificial Neural Network.

Citations

IRE Journals:
Obi-Obuoha Abiamamela , Mubaraq Onipede , Owa Olorunsola Jason "A Comparative Analysis of Machine Learning Algorithms for USD/EUR Foreign Exchange Rate Forecasting" Iconic Research And Engineering Journals Volume 8 Issue 6 2024 Page 443-455

IEEE:
Obi-Obuoha Abiamamela , Mubaraq Onipede , Owa Olorunsola Jason "A Comparative Analysis of Machine Learning Algorithms for USD/EUR Foreign Exchange Rate Forecasting" Iconic Research And Engineering Journals, 8(6)