---
title: New Forecasting Methods for an Old Problem: Predicting 147 Years of Systemic Financial Crises
url: https://www.ml-quant.com/papers/ssrn/7614659/
site: ML-Quant (https://www.ml-quant.com)
updated: 2026-10-16
license: Summaries CC BY 4.0; links go to the original sources
index: https://www.ml-quant.com/llms.txt
identifier: SSRN 7614659
source_url: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7614659
featured: 2026-10-16
citations: unknown
topic: Econometrics & Forecasting
---


# New Forecasting Methods for an Old Problem: Predicting 147 Years of Systemic Financial Crises

Random forests and gradient boosting forecast financial crises across 147 years and 17 countries with 77 percent accuracy, using real-sector predictors as the dominant source.

- Source: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7614659
- Identifier: SSRN 7614659
- Released: 2026-10-15
- First featured: Quant Letter No. 135 (2026-10-16): https://www.ml-quant.com/issues/2026-10-16/
- Citations (Semantic Scholar): not tracked
- Published in: not yet
- Topic: Econometrics & Forecasting
- Authors: Emile du Plessis, Ulrich Fritsche

## Related

- [S&P 500 Trend Prediction](https://www.ml-quant.com/papers/arxiv/2412.11462/): The study uses machine learning to predict S&P 500 trends, concluding that KNN is best for short-term predictions and XGBoost for long-term forecasts.
- [Blending gradient boosted trees and neural networks for point and probabilistic forecasting of hierarchical time series](https://www.ml-quant.com/papers/doi/10-1016-j-ijforecast-2022-01-001/): The paper outlines a successful method for point and probabilistic forecasting using a mix of machine learning models, as demonstrated in the M5 Competition, highlighting the significance of diverse models and careful validation example selection.
- [Generalized Autoregressive Score Trees and Forests](https://www.ml-quant.com/papers/arxiv/2305.18991/): Proposal to improve GAS model forecasts by localizing parameters using decision trees and random forests, outperforming baseline model in empirical analyses.
- [RealTime PV Power Forecasting](https://www.ml-quant.com/papers/ssrn/5241274/): The XGBoost model, using historical weather and PV output data, offers more precise ultrashort-term PV power predictions than the SVR model, contributing to grid stability.
- [Economic Growth Forecasting in Sverdlovsk Region](https://www.ml-quant.com/papers/repec/aiy-jnjaer-v-23-y-2024-i-3-p-674-695/): Machine learning, particularly the random forest model, is more effective in predicting the Gross Regional Product growth of Russia's Sverdlovsk region than traditional models, according to a study.
- [Stock Price Forecasting](https://www.ml-quant.com/papers/ssrn/4906691/): The research uses machine learning to predict stocks that will not have negative returns next year, recommending a Boglehead investment approach, with XGBoost providing the best results.