A Survey of AI Methods and Evaluation in Financial Trading Research

Summary

This systematic literature review summarizes research on artificial intelligence in financial trading published from 2015 to 2023. It reports screening 143 papers across multiple research databases and compares the markets and assets studied, analytical approaches, AI methods, and evaluation measures. Stocks, foreign exchange, and cryptocurrency appear prominently in the surveyed work. Technical analysis is more common than fundamental analysis, while deep learning and reinforcement learning are frequently used; hybrid methods and recurrent neural networks also feature in the review.

The review describes prediction metrics such as error and classification measures alongside investment measures including risk-adjusted returns and drawdown. It identifies data quality, high-frequency data handling, and changing economic conditions as ongoing challenges, and recommends more attention to risk control, crisis detection, automated trading, and combinations of fundamental and technical information. These findings map the reviewed literature rather than establish that AI methods reliably outperform. The summary provides limited detail about search inclusion rules, study quality, or comparability across different markets and performance tests, so its reported prevalence and conclusions should be read as a survey of published work.

Key ideas

• The review examines 143 studies of AI applications in financial trading from 2015 to 2023.
• Stocks, foreign exchange, and cryptocurrency are among the most studied markets in the surveyed literature.
• Technical analysis appears more often than fundamental analysis in the reviewed papers.
• Deep learning and reinforcement learning are common approaches, often combined in hybrid models.
• Forecast accuracy measures and portfolio performance measures assess different aspects of model quality.
• Data quality, market change, risk control, and crisis detection remain challenges for research.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.