Performance Evaluation of Classical and Deep Learning Methods in Diabetes Prediction
Jambura Journal of Biomathematics (JJBM) · 13.06.2026 · 7 (2) · 274-290
DOI: 10.37905/jjbm.v7i2.62 ↗Kurz gefasst
Fachartikel in der Zeitschrift Jambura Journal of Biomathematics (JJBM), online erschienen am 13.06.2026. Verfasst von Buğçe Tatlıcıoğlu. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).
Themen: Diabetes mellitus
Abstract (Originalsprache)
This study presents a comparative evaluation of three approaches for forecasting a diabetes complications dynamical model: the classical fourth-order Runge–Kutta method (RK4), the multiplicative Runge–Kutta method (MRK4), and a Multilayer Perceptron (MLP) trained as a surrogate predictor. RK4 and MRK4 are used to numerically simulate the model, while the MLP is trained on trajectories generated by MRK4. Performance is assessed using mean squared error (MSE), root mean squared error (RMSE), and mean absolute percentage error (MAPE). The results highlight the advantages and limitations of each approach in capturing the model dynamics and provide guidance on when numerical solvers or learning-based surrogates may be preferable in diabetes modeling.
Zitieren
APA
Tatlıcıoğlu, B. (2026). Performance Evaluation of Classical and Deep Learning Methods in Diabetes Prediction. Jambura Journal of Biomathematics (JJBM), 7(2), 274-290. https://doi.org/10.37905/jjbm.v7i2.62
Vancouver
Tatlıcıoğlu B. Performance Evaluation of Classical and Deep Learning Methods in Diabetes Prediction. Jambura Journal of Biomathematics (JJBM). 2026;7(2):274-290. doi:10.37905/jjbm.v7i2.62
BibTeX
@article{Tatlcolu2026,
author = {Tatlıcıoğlu, Buğçe},
title = {Performance Evaluation of Classical and Deep Learning Methods in Diabetes Prediction},
journal = {Jambura Journal of Biomathematics (JJBM)},
year = {2026},
volume = {7},
number = {2},
pages = {274--290},
doi = {10.37905/jjbm.v7i2.62}
}
EUVIMED