ARCHIVES
VOL. 2, ISSUE 4 (2026)
Explainable ensemble machine learning for diabetes prediction and risk assessment
Authors
Gaith Rjoub
Abstract
Diabetes mellitus is a major chronic metabolic disorder that requires
early identification and appropriate risk assessment to reduce the likelihood
of severe complications. The increasing availability of clinical, demographic,
behavioral, and physiological data has created opportunities for machine
learning-based predictive systems capable of identifying complex patterns
associated with diabetes. However, the use of highly complex machine learning
models in healthcare is often challenged by limited interpretability, making it
difficult for healthcare professionals and other stakeholders to understand why
a particular prediction has been generated. This study presents an explainable
ensemble machine learning framework for diabetes prediction and risk assessment
that combines multiple predictive algorithms with explainable artificial
intelligence techniques. The proposed framework integrates ensemble classifiers
such as Random Forest, Gradient Boosting, and XGBoost with feature selection
and SHAP-based model interpretation. Patient characteristics, including
demographic information, anthropometric measurements, clinical indicators, and
relevant behavioral factors, can be used as predictive variables. The
methodology involves data preprocessing, missing-value treatment, feature
selection, model development, ensemble learning, performance evaluation, and
explainability analysis. Model performance can be assessed using accuracy,
precision, recall, F1-score, receiver operating characteristic area under the
curve (ROC-AUC), and calibration measures. SHAP analysis can subsequently be
used to determine the importance and direction of individual predictors. Recent
research demonstrates that stacking ensemble models can achieve strong
predictive performance while SHAP can provide interpretable information about
important diabetes risk factors (Resma et al., 2026) [3].
Earlier research has similarly demonstrated the feasibility of combining
ensemble classification with SHAP explanations for diabetes prediction (Khan et
al., 2022) [1]. The proposed framework therefore emphasizes both
predictive performance and transparency, providing a foundation for developing
trustworthy machine learning systems for diabetes risk assessment.
Nevertheless, external validation, calibration, and clinical assessment remain
necessary before such models can be considered for real-world healthcare applications.
Download
Pages:4-8
How to cite this article:
Gaith Rjoub "Explainable ensemble machine learning for diabetes prediction and risk assessment". World Journal of Current Research, Vol 2, Issue 4, 2026, Pages 4-8
Pdf, Certificate, Cover Page, Index Page, Editorial Page, Accept Letter
Please enter the email address corresponding to this article submission.
