An AI-based hyperautomation framework for algorithmic trading: forecasting automated trading strategy profitability from market data
Abstract
An “always-on” Expert Advisor can trade even in poor market conditions. This study adds a simple decision layer that turns the EA on or off each week. Using weekly data, AI models predict whether the EA is likely to make money next week and activate it only when the outlook is positive. Logistic Regression, XGBoost, and LightGBM are compared with LSTM-based models. Tests on 2020–2025 data for five instruments and MT5 backtests show higher Sharpe Ratio and Profit Factor, with lower maximum drawdown, than keeping the EA always on. Tree models work best for major FX; LSTMs suit gold.
Fields of Study & Subjects
Stock Market Forecasting Methods
Forecasting Techniques and Applications
Customer churn and segmentation
Publication details
- Journal / source
- Enterprise Information Systems
- Publisher
- Informa UK Limited
- Volume
- 20
- Issue
- 5-6
- Publication date
- 18 February 2026
- Language
- EN