PREDICTIVE POWER OF MACHINE LEARNING MODELS IN FOREX MARKET: A COMPARATIVE STUDY
PREDICTIVE POWER OF MACHINE LEARNING MODELS IN FOREX MARKET: A COMPARATIVE STUDY

Summary/Abstract: In this paper, machine learning models for Forex prediction,evaluating traditional ensemble methods (Random Forest, XGBoost,LightGBM) against specialized time series models (Prophet, Arima, LSTM)across multiple currency pairs are compared. Performance assessment usesboth statistical metrics (RMSE, MAE, directional accuracy) and tradingmeasures (Sharpe ratio, maximum drawdown) across different marketconditions. It is shown that ensemble methods excel with rich feature setswhile time series models better capture temporal patterns. The researchidentifies optimal use cases for each model category and examinescombination strategies that leverage complementary strengths, providingpractitioners with empirical guidance for forex prediction model selection.

• Details
• Contents