EUVIMEDEUVIMED entdecken
Crossref

RANDOM FOREST-BASED CREDIT RISK CLASSIFICATION FOR DOST-SETUP ENTERPRISES: A PRELIMINARY DECISION-SUPPORT MODEL

Gerald S. Olloves, Kennet G. Cuarteros

Far East Journal of Theoretical Statistics · 2026 · 70 (4) · 659-670

DOI: 10.17654/0972086326031 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Far East Journal of Theoretical Statistics, 2026. Verfasst von Gerald S. Olloves und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Government technology-assistance programs require consistent methods for identifying enterprises at risk of repayment difficulty. This study developed and evaluated a Random Forest classifier for 489 projects supported by the Department of Science and Technology-Small Enterprise Technology Upgrading Program (DOST-SETUP) in Northern Mindanao from 2000 to 2024. The data included 281 Non-Default and 208 Default projects and 18 pre-intervention predictors. To approximate prospective use and prevent later projects from informing earlier predictions, models were trained on projects obligated through \(2020(n=326)\) and tested on projects from 2021-2024 \((n=163)\). On the temporal test set, Random Forest achieved 61.3% accuracy (95% CI: 54.0%-68.1%), Default recall of \(42.9 \%(30.2 \%-55.6 \%)\), and an AUC of 0.625 (0.535-0.711). Logistic Regression performed better, with 67.5% accuracy and an AUC of 0.718, whereas the Decision Tree achieved 49.7% accuracy and an AUC of 0.535. Five-fold cross-validation within the pre-2021 training period produced a Random Forest AUC of 0.621 ± 0.053. The six leading Random Forest predictors were used in a supplementary points-based scorecard; its historical risk gradient was descriptive because construction and assessment used overlapping development data. Existing SETUP administrative data therefore contain modest predictive signal, but the Random Forest does not support standalone automated screening. It is best regarded as an exploratory decision-support model requiring temporal replication, external validation, calibration, and richer business and market predictors.

Zitieren

APA

Olloves, G. S., & Cuarteros, K. G. (2026). RANDOM FOREST-BASED CREDIT RISK CLASSIFICATION FOR DOST-SETUP ENTERPRISES: A PRELIMINARY DECISION-SUPPORT MODEL. Far East Journal of Theoretical Statistics, 70(4), 659-670. https://doi.org/10.17654/0972086326031

Vancouver

Olloves GS, Cuarteros KG. RANDOM FOREST-BASED CREDIT RISK CLASSIFICATION FOR DOST-SETUP ENTERPRISES: A PRELIMINARY DECISION-SUPPORT MODEL. Far East Journal of Theoretical Statistics. 2026;70(4):659-670. doi:10.17654/0972086326031

BibTeX
@article{Olloves2026,
  author = {Olloves, Gerald S. and Cuarteros, Kennet G.},
  title = {RANDOM FOREST-BASED CREDIT RISK CLASSIFICATION FOR DOST-SETUP ENTERPRISES: A PRELIMINARY DECISION-SUPPORT MODEL},
  journal = {Far East Journal of Theoretical Statistics},
  year = {2026},
  volume = {70},
  number = {4},
  pages = {659--670},
  doi = {10.17654/0972086326031}
}
Kostenlos starten →