Navigating Financial Instability: An Interpretable Paradigm for Multi-Label Crisis Prediction
Bibliographic record
Abstract
We present a rigorous methodology for applying Elastic Weight Consolidation (EWC) to the continual prediction of banking, systemic, currency and inflation crises. Traditional models, trained on earlier regimes, are prone to catastrophic forgetting when updated on new regimes for financial crisis prediction that is a sequential, non-stationary learning problem with data arriving across time and heterogeneous jurisdictions. EWC mitigates this by constraining important parameters, quantified via a diagonal-Fisher-information approximation, to remain near previously learnt optima while permitting flexibility in less critical directions. We formalise the Bayesian derivation underpinning EWC, present the practical diagonalFisher estimator, and integrate the method into a reproducible training pipeline for multi-label crisis prediction. Our protocol uses representative task partitions, rigorous Fisher estimation, lambda scheduling, and a suite of evaluation metrics to compare performance with baseline traditional models. We provide ablation studies for Fisher estimation, penalty aggregation, and sequential walk-forward evaluation, and discuss limitations and interpretability for policy use. Our experimental model is tested for 2 years prior financial crisis prediction on Harvard Business School's global crisis database for 42 years and 70 countries, and it demonstrates ROC-AUC of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 3. 1 2 \%}$</tex> in inflation crises, <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{8 3. 5 3 \%}$</tex> in systemic crises, 77.63% in currency crises and 67.51% in banking crises predictions, outperforming the traditional models' results. We also assess the sensitivity of EWC to its core hyperparameters for real-world application. By integrating SHAP, this work aims to bridge probabilistic continual-learning theory and operational early-warning systems, making models both interpretable and robust in dynamic financial environments.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.005 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".