What is it about?
This review maps how AI and machine learning are used to predict financial markets stocks, crypto, commodities and FX. From 100 studies, it shows which models are most common (e.g., LSTM/GRU and XGBoost), how accuracy is typically measured, and where predictions tend to fail in practice. It also highlights recurring data and validation challenges.
Featured Image
Why is it important?
Financial forecasting research is exploding, but it is hard to see what truly works. Using a PRISMA-based systematic review of 100 papers, we identify go-to models and metrics, then spotlight gaps that limit real-world impact: forecasts are rarely tested inside trading strategies, metrics like MAPE can mislead in volatile markets, and many studies don’t check performance across regimes. We outline a clear agenda: interpretability, strategy-level evaluation, and volatility-aware validation.
Perspectives
What I found most striking is the gap between “good forecast errors” and practical usefulness. Across asset classes, many papers optimize RMSE/MAPE but stop short of asking the investor’s question: does this translate into a robust strategy after costs and changing regimes? Writing the review reinforced my view that the next wave should focus less on ever-new architectures and more on transparent models, regime testing, and decision-focused evaluation.
Tibor Bareith
ELTE Centre for Economic and Regional Studies Hungary
Read the Original
This page is a summary of: Navigating AI-Driven Financial Forecasting: A Systematic Review of Current Status and Critical Research Gaps, Forecasting, July 2025, MDPI AG,
DOI: 10.3390/forecast7030036.
You can read the full text:
Contributors
The following have contributed to this page
