Saved in:
| Main Author: | |
|---|---|
| Format: | Recurso digital |
| Language: | |
| Published: |
Zenodo
2026
|
| Online Access: | https://doi.org/10.5281/zenodo.18919578 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Table of Contents:
- <div> <div>Diabetes needs to be detected early and accurately to receive medical attention early and reduce long-term complications. This paper will offer a hybrid machine learning model that can be explained to predict a risk of diabetes and offer clinical decision support using secondary clinical data. The proposed framework integrates stacking ensemble with the use of a Random Forest, Gradient Boosting and Multi-Layer Perceptron classifiers and a logistic regression meta-learner. This is done by adding to the predictive layer a similarity based reasoning mechanism and a rule based clinical validation layer to enhance the clinical reliability of the model. 10-fold stratified cross-validation is used to evaluate the model performance. Stacking ensemble showed the best accuracy of 0.94, precision of 0.93, recall of 0.92, F1-score of 0.93 and ROC-AUC of 0.97, which are superior to the other baseline models. SHAP based both global and local explanations provide explanability, as it is easy to determine key clinical feaures, which influence predictions. The results indicate that the proposed framework provides valid, decipherable and clinically practicable diabetes risk measurements to be utilized in decision support programs.</div> </div>