An integrated hybrid soft voting ensemble ai model of machine learning and deep learning models for diabetes prediction

Islam, Md Ziarul and Hassan, Mohd Khairul Azmi and Amir Hussin, Amir 'Aatieff and Ikram Kays, H M (2026) An integrated hybrid soft voting ensemble ai model of machine learning and deep learning models for diabetes prediction. International Journal of Software Engineering and Computer Systems, 11 (2). pp. 160-175. ISSN 2289-8522 E-ISSN 2180-0650

PDF - Published Version Restricted to Registered users only Download (4MB) | Request a copy

Abstract

The goal of the study is to make a hybrid prediction model that uses both machine learning and deep learning methods to make diabetes predictions more accurate, generalizable, and strong. It combines ML and DL models, fixes class imbalance in medical datasets, and tests performance on several datasets, such as the Pima Indians Diabetes Dataset and the LMCH dataset, to see how well it works in real-life healthcare. The ML Ensemble, which included RF, LR, and XGBoost, and the DL Ensemble, which included CNN, FNN, and ENN, were the two stacked ensembles used in the study. Soft voting was used to aggregate the results in order to improve the accuracy of the predictions. In order to prepare the structured medical data, we employed feature preprocessing techniques and the Synthetic Minority Over-sampling Technique (SMOTE). Cross-validation was used to ensure that the results were good and to prevent them from being overly specific. The performance was compared to independent models and standard methods. The ensemble hybrid AI model performed better than the traditional ML and DL models. Its best performance metrics were Accuracy of 98.89%, Precision of 98.99%, Recall of 87.07%, F1-score of 92.05%, ROC-AUC of 92.48%, and Cohen's Kappa of 84.96%. This shows that it was better at making generalizations and working with datasets that weren't balanced. The stacking of ensembles with soft voting combines machine learning and deep learning models to improve diabetes prediction performance and fix problems with class imbalance in medical datasets. The model's ability to be used in the real world and its ability to be generalized show that it could be used to find diabetes early and accurately, which could help with preventive healthcare strategies

Item Type:	Article (Journal)
Uncontrolled Keywords:	Diabetes prediction, Machine learning, Deep learning, Soft voting classifier, AI-Based medical diagnosis, Hybrid ensemble learning
Subjects:	T Technology > T Technology (General)
Kulliyyahs/Centres/Divisions/Institutes (Can select more than one option. Press CONTROL button):	Kulliyyah of Information and Communication Technology > Department of Information System Kulliyyah of Information and Communication Technology > Department of Information System
Depositing User:	Dr Mohd Khairul Azmi Hassan
Date Deposited:	16 Jun 2026 12:27
Last Modified:	16 Jun 2026 12:27
Queue Number:	2026-06-Q3666
URI:	http://irep.iium.edu.my/id/eprint/129281

Actions (login required)

View Item