AI summaryⓘ
The authors created a large new dataset called V4FinBench to help predict when companies might go bankrupt, which is challenging because bankrupt companies are very rare. This dataset contains over a million company-year records from four Central European countries, with many financial and non-financial features and several time-based prediction goals. They tested existing machine learning methods and found that one method, TabPFN, works well for longer-term predictions when adjusted for class imbalance, while a large language model (Llama-3-8B) performed worse than traditional methods. The authors also showed that models trained on their dataset can improve predictions on a different American dataset, meaning their data and methods capture useful financial distress signals. They made the dataset publicly available to encourage better research in this important area.
corporate bankruptcy predictionclass imbalancetabular datafinancial distressmachine learningTabPFNgradient boostingROC-AUCF1-scorebenchmark dataset
Authors
Marcin Kostrzewa, Sebastian Tomczak, Roman Furman, Anna Poberezhna, Michał Furgała, Oleksii Furman, Maciej Zięba
Abstract
Corporate bankruptcy prediction is a high-stakes financial task characterized by severe class imbalance and multi-horizon forecasting demands. Public datasets supporting it remain scarce and small: widely used free benchmarks contain between 6,000 and 80,000 company-year observations, while larger resources are behind subscription paywalls. To address this gap, we introduce V4FinBench, a benchmark of over one million company-year records from the Visegràd Group (V4) economies (2006-2021), with 131 financial and non-financial features, six prediction horizons, and a composite distress criterion jointly capturing solvency, profitability, and liquidity deterioration. V4FinBench is designed to support the evaluation of tabular and foundation-model methods under realistic class imbalance, with positive rates between 0.19% and 0.36%. We provide reference evaluations of standard tabular baselines, finetuned TabPFN, and QLoRA-finetuned Llama-3-8B. With imbalance-aware finetuning, TabPFN matches or exceeds gradient boosting at longer time horizons on both $F_1$-score and ROC-AUC. In contrast, Llama-3-8B trails gradient boosting on ROC-AUC at every horizon and is generally weaker on $F_1$-score, with the gap widening sharply beyond the immediate horizon. In an external evaluation on the American Bankruptcy Dataset, the V4FinBench-finetuned TabPFN checkpoint improves over vanilla TabPFN, suggesting that adaptation captures transferable financial-distress structure rather than only V4-specific patterns. V4FinBench is publicly released to support further evaluation and development of prediction methods on realistic financial data.