EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Bridging Gaps in Supervised Machine Learning: A Standardized Empirical Benchmarking Study on Real-World Tasks

Muhammad Tella, Kabiru Ibrahim Musa, Mahmud Ahmed Usman

International Journal of Scientific Research in Computer Science and Engineering · 31.12.2025 · 13 (6) · 1-7

DOI: 10.26438/ijsrcse.v13i6.792 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Scientific Research in Computer Science and Engineering, online erschienen am 31.12.2025. Verfasst von Muhammad Tella und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Supervised machine learning (SML) has direct application on various real-world problems and there is need for frameworks that are not only accurate but also robust and standardized in order to guide algorithm selection. Previous studies have compared classifiers across domains such as ecology and healthcare. However, the comparisons have certain shortcomings such as reliance on heterogeneous datasets, inconsistent evaluation metrics and limited ensemble strategies thereby weakening the reliability and generalizability of findings. This study proposes a framework that presents a systematic empirical benchmarking of some selected single classifiers and ensemble methods across multiple real-world datasets that are publicly available. Training and validation of models employed uniform preprocessing, hyperparameter tuning with 10-fold cross-validation, and three consistent evaluation metrics, namely accuracy, F1-score and AUC, aimed at ensuring fair and reproducible comparisons. Computational efficiency of models and their interpretability were also evaluated in which the ensemble methods, particularly stacking, were found to demonstrate performance that significantly exceed those of single classifiers with the ensemble methods achieving the highest accuracy (92.1%), F1-score (0.91), and AUC (0.96) across datasets, with statistically significant improvements confirmed via paired t-tests. Despite the fact that stacking displayed a highly improved predictive performance and stability, it incurs a higher training time which can be considered as a trade-off. These finding underscore the value of standard benchmarking frameworks in supervised learning and provide actionable guidance for selecting models that balance accuracy, interpretability, and computational efficiency. The study contributes both theoretically (through supporting statistical learning theory) and practical recommendations for deploying machine learning models in diverse real-world applications.

Zitieren

APA

Tella, M., Musa, K. I., & Usman, M. A. (2025). Bridging Gaps in Supervised Machine Learning: A Standardized Empirical Benchmarking Study on Real-World Tasks. International Journal of Scientific Research in Computer Science and Engineering, 13(6), 1-7. https://doi.org/10.26438/ijsrcse.v13i6.792

Vancouver

Tella M, Musa KI, Usman MA. Bridging Gaps in Supervised Machine Learning: A Standardized Empirical Benchmarking Study on Real-World Tasks. International Journal of Scientific Research in Computer Science and Engineering. 2025;13(6):1-7. doi:10.26438/ijsrcse.v13i6.792

BibTeX
@article{Tella2025,
  author = {Tella, Muhammad and Musa, Kabiru Ibrahim and Usman, Mahmud Ahmed},
  title = {Bridging Gaps in Supervised Machine Learning: A Standardized Empirical Benchmarking Study on Real-World Tasks},
  journal = {International Journal of Scientific Research in Computer Science and Engineering},
  year = {2025},
  volume = {13},
  number = {6},
  pages = {1--7},
  doi = {10.26438/ijsrcse.v13i6.792}
}
Kostenlos starten →