(Taylor & Francis): Evaluates 497 trading rules across 10 currencies . Summary Table of Model Performance Model Type Primary Strength Reported Accuracy Multi-Layer Perceptron (MLP) High ROI & profitability Up to 86.21% SARIMA / Prophet Strong directional accuracy Up to 86.21% Logistic Regression Stability in hourly forecasting DPCN (Deep Predictive Coding) Short-term trend prediction >99% (in specific conditions)
For in-depth study, these papers provide comprehensive frameworks: ACCURATE FOREX SIGNALS
models, such as Event-driven LSTM and TradingBERT, are increasingly used to extract signals by analyzing high-frequency data . (Taylor & Francis): Evaluates 497 trading rules across
Recent studies emphasize that high prediction accuracy does not always equate to higher ROI due to market noise and execution lag . Key findings from contemporary papers include: these papers provide comprehensive frameworks: models
(arXiv): Details the implementation of an AI-based system for EUR-USD using hybrid variable sets .
: A study on hourly EUR/USD directional forecasting achieved 58.5% mean out-of-sample accuracy , emphasizing that "temporal misalignment" between a signal and a trade often ruins profitability . Machine Learning Performance :
Before applying these findings, it is highly recommended to (use a demo account) to verify signal accuracy in real-time without risking capital .