EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

AN EXPLAINABLE HYBRID MACHINE LEARNING FRAMEWORK FOR CASH-FLOW-CONDITIONED MULTI-HORIZON LIQUIDITY RISK EARLY WARNING: AN SME-ORIENTED BENCHMARK STUDY

XinRan Yue, RuFeng Zhu

World Journal of Management Science · 01.01.2024 · 2 (1) · 63-72

DOI: 10.61784/wms4120 ↗

Kurz gefasst

Fachartikel in der Zeitschrift World Journal of Management Science, online erschienen am 01.01.2024. Verfasst von XinRan Yue und 1 weiteren Autor:innen. Die Arbeit wurde bisher 19-mal zitiert (Crossref). Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Small and medium-sized enterprises (SMEs) are particularly exposed to liquidity shocks, yet public, transaction-level SME cash-flow panels suitable for reproducible forecasting remain scarce. This paper therefore studies a narrower but operationally related task: cash-flow-conditioned, multi-horizon liquidity-risk early warning. A real public benchmark containing 43,405 financial-statement observations, 64 ratios, missing values, and bankruptcy labels one to five years ahead is used as an evidence base. The proposed framework combines three views: an L2-regularized logistic model, a LightGBM model using all ratios, and a dedicated LightGBM expert using 23 cash-flow and liquidity ratios. Non-negative horizon-specific weights are selected on a calibration split in log-odds space, followed by isotonic probability calibration, a recall-constrained warning threshold, and SHAP-based feature and semantic-group explanations. Across five held-out horizons, the calibrated hybrid achieved mean ROC-AUC 0.914, PR-AUC 0.617, and Brier score 0.0260. The uncalibrated ensemble had higher ranking performance (PR-AUC 0.663) but materially worse probability error (Brier 0.0424), exposing a genuine calibration-discrimination trade-off. The cash-flow expert received positive weight at the two- and five-year horizons, while the full-feature tree dominated the remaining horizons; the PR-AUC gain over LightGBM-all was not statistically significant (one-sided Wilcoxon p=0.50). The study supplies executable code, data hashes, result tables, figures, and verified references. Because the benchmark does not identify firms by a legal SME definition and is not a cash-flow time series, the findings support an SME-oriented methodology rather than a claim of direct SME field validation.

Zitieren

APA

Yue, X., & Zhu, R. (2024). AN EXPLAINABLE HYBRID MACHINE LEARNING FRAMEWORK FOR CASH-FLOW-CONDITIONED MULTI-HORIZON LIQUIDITY RISK EARLY WARNING: AN SME-ORIENTED BENCHMARK STUDY. World Journal of Management Science, 2(1), 63-72. https://doi.org/10.61784/wms4120

Vancouver

Yue X, Zhu R. AN EXPLAINABLE HYBRID MACHINE LEARNING FRAMEWORK FOR CASH-FLOW-CONDITIONED MULTI-HORIZON LIQUIDITY RISK EARLY WARNING: AN SME-ORIENTED BENCHMARK STUDY. World Journal of Management Science. 2024;2(1):63-72. doi:10.61784/wms4120

BibTeX
@article{Yue2024,
  author = {Yue, XinRan and Zhu, RuFeng},
  title = {AN EXPLAINABLE HYBRID MACHINE LEARNING FRAMEWORK FOR CASH-FLOW-CONDITIONED MULTI-HORIZON LIQUIDITY RISK EARLY WARNING: AN SME-ORIENTED BENCHMARK STUDY},
  journal = {World Journal of Management Science},
  year = {2024},
  volume = {2},
  number = {1},
  pages = {63--72},
  doi = {10.61784/wms4120}
}
Kostenlos starten →