@article {10.3844/jcssp.2026.111.120,
article_type = {journal},
title = {Comparative Analysis of Neural Network Models for Predicting EUR/USD Direction: An Empirical Study},
author = {Mohamed, El Badaoui and Brahim, Raouyane and Samira, El Moumen and Mostafa, Bellafkih},
volume = {22},
number = {1},
year = {2026},
month = {Feb},
pages = {111-120},
doi = {10.3844/jcssp.2026.111.120},
url = {https://thescipub.com/abstract/jcssp.2026.111.120},
abstract = {This paper presents a rigorous comparative analysis of six feedforward neural network models for predicting the directional movement of the EUR/USD currency pair. The evaluated models include the Learning Vector Quantization, Cascade Neural Network, Feedforward Neural Network, Single Layer Perceptron, Multi-Layer Perceptron, and Radial Basis Function network. Utilizing daily historical data from April 2009 to May 2024, each model was trained and optimized under uniform conditions on a rich feature set derived from a diverse pool of technical indicators. Model performance was comprehensively evaluated using a suite of metrics, including accuracy, MSE, MAE, R², balanced accuracy, F1-score, precision, recall, and the sharpe ratio. The Cascade Neural Network consistently demonstrated superior performance, achieving a validation accuracy of 74.8, a balanced accuracy of 74.8, and a validation F1-score of 75.44%. By establishing a robust performance baseline for these foundational architectures, this study highlights the significant potential of neural networks in forex forecasting and provides critical insights into their respective strengths and weaknesses. The findings serve as a guide for future research and practical applications in financial market analysis, particularly in the development of more advanced predictive systems.},
journal = {Journal of Computer Science},
publisher = {Science Publications}
}