EUVIMEDEUVIMED entdecken
Crossref

An empirical analysis of Bayesian-optimized boosting ensembles for medical relief demand forecasting

Roman Villones, Jonilo Mababa, Jovy Jay Cabrera, Jaime Pulumbarit

International Journal of Science, Technology, Engineering and Mathematics · 01.06.2026 · 6 (2) · 1

DOI: 10.53378/ijstem.353345 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Science, Technology, Engineering and Mathematics, online erschienen am 01.06.2026. Verfasst von Roman Villones und 3 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

In humanitarian logistics, the planning of medical relief supply is sensitive to accurate demand forecasting due to its intermittent and volatile demand pattern that constrain the usefulness of conventional statistical techniques. This paper provides an empirical analysis of the increase in ensemble learning models that are optimized through Bayesian searching of hyperparameters to predict demand of medical relief. Using the Team Data Science Process (TDSP) framework, a quantitative methodology is used to evaluate the performance of the model based on predictive accuracy and robustness performance with extreme values, and generalizability on a real-world dataset in the National Capital Region (NCR), Philippines. There are five ensemble models such as AdaBoost, CatBoost, Gradient Boosting, LightGBM, and XGBoost. Findings indicate that Bayesian optimization produces a tangible performance gain, especially on Gradient Boosting and LightGBM. The CatBoost model produces the lowest RMSE, WAPE, and MASE, and the most consistent cross-validation results, which means that it is more accurate and stable in the model tested. On the contrary, XGBoost and AdaBoost exhibits relatively poorer performance and low robustness. Although the results illustrate the efficiency of optimized boosting ensembles to handle complex and irregular demand shapes, the research study is limited by the coverage of datasets, possible temporal leakage and lack of actual deployment. Thus, it is possible to arrive at conclusions only based on the assessed data and experimental conditions.

Zitieren

APA

Villones, R., Mababa, J., Cabrera, J. J., & Pulumbarit, J. (2026). An empirical analysis of Bayesian-optimized boosting ensembles for medical relief demand forecasting. International Journal of Science, Technology, Engineering and Mathematics, 6(2), 1. https://doi.org/10.53378/ijstem.353345

Vancouver

Villones R, Mababa J, Cabrera JJ, Pulumbarit J. An empirical analysis of Bayesian-optimized boosting ensembles for medical relief demand forecasting. International Journal of Science, Technology, Engineering and Mathematics. 2026;6(2):1. doi:10.53378/ijstem.353345

BibTeX
@article{Villones2026,
  author = {Villones, Roman and Mababa, Jonilo and Cabrera, Jovy Jay and Pulumbarit, Jaime},
  title = {An empirical analysis of Bayesian-optimized boosting ensembles for medical relief demand forecasting},
  journal = {International Journal of Science, Technology, Engineering and Mathematics},
  year = {2026},
  volume = {6},
  number = {2},
  pages = {1},
  doi = {10.53378/ijstem.353345}
}
Kostenlos starten →