What is it about?
Financial stress can build quickly, while many traditional early-warning systems rely on indicators that are updated too slowly to capture fast-moving changes in markets. This study develops a high-frequency framework for monitoring systemic financial stress in Europe using daily information from financial markets, banks, interest rates, exchange rates, commodities, volatility, and sentiment indicators. We compare statistical and machine-learning models under a strictly chronological forecasting design that prevents future information from leaking into past predictions. The results show that systemic stress is most effectively monitored over short horizons, particularly around five business days ahead. A relatively transparent logistic model using measures related to European systemic stress and market volatility performs especially well, while more complex nonlinear machine-learning models do not consistently outperform it. Related keywords: systemic financial stress systemic risk financial stability early-warning system machine learning financial crisis prediction high-frequency financial data CISS Composite Indicator of Systemic Stress VIX rare-event prediction financial stress monitoring central banking macro-financial risk interpretable machine learning chronological validation time-series forecasting
Featured Image
Photo by Sortter on Unsplash
Why is it important?
Financial crises and periods of systemic stress can develop faster than conventional macroeconomic monitoring systems can respond. This research shows that daily financial and sentiment data can provide useful short-horizon signals of increasing systemic stress in Europe. The framework is particularly relevant for central banks, financial regulators, risk managers, and researchers developing early-warning systems. Importantly, the findings also show that greater model complexity does not automatically produce better monitoring: a transparent and interpretable model can remain highly competitive with nonlinear machine-learning approaches. This makes the framework potentially valuable for real-world financial-stability monitoring, where predictive performance, interpretability, and timely decision support are all important.
Perspectives
A key lesson from this study is that the most sophisticated algorithm is not necessarily the most useful one for financial-stability monitoring. When systemic stress is rare and market conditions change quickly, disciplined validation, appropriate stress definitions, and informative high-frequency predictors can matter as much as model complexity. Our results suggest that short-horizon monitoring should be treated as a decision-support problem rather than as an attempt to predict financial crises far in advance. We hope the framework encourages further research on high-frequency early-warning systems, alternative definitions of systemic stress, cross-country applications, and interpretable machine learning for financial stability.
Prof. Afshin Ashofteh
Universidade Nova de Lisboa
Read the Original
This page is a summary of: A machine learning framework for short-horizon monitoring of systemic financial stress in Europe, Decision Analytics Journal, September 2026, Elsevier,
DOI: 10.1016/j.dajour.2026.100742.
You can read the full text:
Resources
A machine learning framework for short-horizon monitoring of systemic financial stress in Europe
Diachkov, D., & Ashofteh, A. (2026). A machine learning framework for short-horizon monitoring of systemic financial stress in Europe. Decision Analytics Journal, 20, Article 100742. https://doi.org/10.1016/j.dajour.2026.100742 Early detection of systemic financial stress is essential for financial-stability monitoring in environments characterized by rapid market adjustment, nonlinear transmission, and rare stress episodes. This study develops a high-frequency machine-learning framework for short-horizon monitoring of systemic financial stress in Europe using daily financial, banking, interest-rate, exchange-rate, commodity, volatility, and sentiment-related indicators. Systemic financial stress is measured with the European Central Bank’s Composite Indicator of Systemic Stress (CISS), which is transformed into binary stress targets using fixed, rolling-sigma, and rolling-percentile threshold rules. The empirical design evaluates alternative model classes, dataset specifications, feature representations, forecast horizons, and stress definitions within a strictly chronological out-of-sample validation framework that avoids look-ahead bias. Model performance is assessed using rare-event metrics, with primary emphasis on precision–recall AUC, recall, and F1 score. The results show that predictive performance is strongest under the fixed CISS threshold and at short horizons, particularly the 5-business-day forecast window. The preferred specification is a transparent logit model using CISS- and VIX-based predictors with level and change features. Sentiment and uncertainty-related indicators also provide meaningful short-horizon information, while broader standalone market blocks perform more moderately. Nonlinear classifiers are competitive in several settings but do not materially dominate the linear benchmark. Robustness checks show that rolling thresholds preserve the qualitative short-horizon pattern but reduce performance and increase specification instability, especially at longer horizons. Overall, the framework is best interpreted as a disciplined decision-analytics tool for high-frequency systemic-stress monitoring rather than as a threshold-invariant long-horizon stress-prediction model.
NOVA Research Portal record
NOVA Research Portal record with manuscript metadata
IEEE 50th Annual Computers, Software, and Applications Conference (COMPSAC)
Diachkov, D., & Ashofteh, A. (2026). High-Frequency Early Warning of Systemic Financial Stress in Europe Using Financial and Non-Financial Data with Machine Learning. In 2026 IEEE 50th Annual Computers, Software, and Applications Conference (COMPSAC) (pp. 2023-2028). (Proceedings of the Annual Computer Software and Applications Conference). IEEE Computer Society. https://doi.org/10.1109/COMPSAC69091.2026.00296
ResearchGate publication page
On Researchgate Diachkov, D., & Ashofteh, A. (2026). A machine learning framework for short-horizon monitoring of systemic financial stress in Europe. Decision Analytics Journal, 20, Article 100742. https://doi.org/10.1016/j.dajour.2026.100742
Contributors
The following have contributed to this page
