EUVIMEDEUVIMED entdecken
Crossref

ADVANCING DATA-DRIVEN PREDICTIVE MODELING: EVALUATING THE PERFORMANCE, ROBUSTNESS, AND PRACTICAL USABILITY OF AUTOMATED MACHINE LEARNING COMPARED TO EXPERT-DESIGNED END-TO-END DATA SCIENCE PIPELINES

Sungho Kim, Md Lutfur Rahman Fahad, Niaz Mahmood Kyoom, Pritom Das, Novera Mahjabin Hossain, Nayem Miah, Mukesh Gogu, Changgi Hong, Ikbal Hossain, Shadia Chowdhury

International Journal of Grid Computing & Applications · 28.03.2026 · 17 (1) · 59-71

DOI: 10.5121/ijgca.2026.17105 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Grid Computing & Applications, online erschienen am 28.03.2026. Verfasst von Sungho Kim und 9 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Predictive modeling has become a foundational component of modern data-driven decisionmaking across business, finance, healthcare, and public policy. Traditionally, predictive modeling pipelines have been designed and optimized by human experts through manual data preprocessing, feature engineering, model selection, and validation. Recent advances in Automated Machine Learning (AutoML) (Hutter, Kotthoff, & Vanschoren, 2019; Zöller & Huber, 2021) promise to automate this end-to-end process, enabling faster model development with reduced human intervention. Despite increasing adoption, there remains limited empirical evidence (Khurana et al., 2018; Truong et al., 2019) evaluating whether AutoML systems can consistently match or replace expert-designed predictive modeling pipelines under realistic constraints. This study presents a comparative evaluation of automated and expert-designed predictive modeling pipelines across multiple tabular datasets. The comparison focuses not only on predictive performance, but also on robustness, efficiency, and practical usability, including interpretability considerations. Using controlled experimental settings with equal data access, evaluation metrics, and computational budgets, we assess the strengths and limitations of each approach. Results are expected to provide evidence-based guidance on when AutoML is sufficient, when human expertise remains essential, and how hybrid strategies may offer the greatest practical value.

Zitieren

APA

Kim, S., Fahad, M. L. R., Kyoom, N. M., Das, P., Hossain, N. M., Miah, N., Gogu, M., Hong, C., Hossain, I., & Chowdhury, S. (2026). ADVANCING DATA-DRIVEN PREDICTIVE MODELING: EVALUATING THE PERFORMANCE, ROBUSTNESS, AND PRACTICAL USABILITY OF AUTOMATED MACHINE LEARNING COMPARED TO EXPERT-DESIGNED END-TO-END DATA SCIENCE PIPELINES. International Journal of Grid Computing & Applications, 17(1), 59-71. https://doi.org/10.5121/ijgca.2026.17105

Vancouver

Kim S, Fahad MLR, Kyoom NM, Das P, Hossain NM, Miah N, et al. ADVANCING DATA-DRIVEN PREDICTIVE MODELING: EVALUATING THE PERFORMANCE, ROBUSTNESS, AND PRACTICAL USABILITY OF AUTOMATED MACHINE LEARNING COMPARED TO EXPERT-DESIGNED END-TO-END DATA SCIENCE PIPELINES. International Journal of Grid Computing & Applications. 2026;17(1):59-71. doi:10.5121/ijgca.2026.17105

BibTeX
@article{Kim2026,
  author = {Kim, Sungho and Fahad, Md Lutfur Rahman and Kyoom, Niaz Mahmood and Das, Pritom and Hossain, Novera Mahjabin and Miah, Nayem and Gogu, Mukesh and Hong, Changgi and Hossain, Ikbal and Chowdhury, Shadia},
  title = {ADVANCING DATA-DRIVEN PREDICTIVE MODELING: EVALUATING THE PERFORMANCE, ROBUSTNESS, AND PRACTICAL USABILITY OF AUTOMATED MACHINE LEARNING COMPARED TO EXPERT-DESIGNED END-TO-END DATA SCIENCE PIPELINES},
  journal = {International Journal of Grid Computing & Applications},
  year = {2026},
  volume = {17},
  number = {1},
  pages = {59--71},
  doi = {10.5121/ijgca.2026.17105}
}
Kostenlos starten →