Label validity before model choice: a leakage-aware evaluation of early-warning prediction in enterprise event logs
Objectives To examine which early-warning claims remain defensible when event-log features, duration thresholds, and development labels are restricted to information available at deployment.Methods Three public logs (188,944 cases; 1,850,318 events) were split chronologically by case arrival into training, calibration, validation, and test sets. Kaplan–Meier duration thresholds used historical training events only; development prefixes and label-availability times preceded the first test arrival. Logistic regression and LightGBM were compared with a GRU benchmark at event-count checkpoints. Uncertainty used case, arrival-month, circular moving-block, and short-window weekly bootstrap schemes.Key findings Eleven of 32 in-scope configurations were evaluable. Loan delay AUC ranged 0.639–0.705, permit repetition AUC 0.530–0.932, and traffic-fine repetition AUC 0.485–0.723. Trajectories describe selected populations, not learning effects; 157 of 562 common permit cases changed labels. Concept additions produced small AUC changes, and the GRU showed no uniform advantage. Moving-block intervals condition on fixed fitted models and therefore do not include model-selection or fitting uncertainty in this final revision.Novelty The study provides an auditable reporting protocol that makes label availability, non-evaluable tasks, checkpoint selection, temporal dependence, and the limits of observational guidance explicit before model comparisons are interpreted.
Authors
- Sabina-Cristiana Necula (ORCID: https://orcid.org/0000-0001-6912-2450)
Institutions
- Alexandru Ioan Cuza University (RO)
Publication Details
- Journal
- Journal of Business Analytics
- Published
- 2026-11-19
- DOI
- https://doi.org/10.1080/2573234x.2026.2762877
- Primary Topic
- Financial Distress and Bankruptcy Prediction
- Type
- article
- Field-Weighted Citation Impact
- 0.00