Using machine learning to monitor financial markets
Keynote speech by Gaston Gelos, Deputy Head, Monetary and Economic Department of the BIS, at the 6th Biennial Conference on Financial Stability, Bank of Mexico, Mexico City, 18 November 2025.
Machine learning and artificial intelligence offer new opportunities to predict market stress and dysfunction, overcoming the limitations of traditional econometric models. BIS research demonstrates that tools like random forests and recurrent neural networks (RNNs) can forecast financial market conditions and identify key drivers of stress. These models improve prediction accuracy by capturing non-linear dynamics and handling large datasets. Additionally, integrating large language models (LLMs) enables supervisors to analyse relevant news and data linked to stress indicators, enhancing monitoring and risk mitigation capabilities.
