Noura Alotaibi, Manal M. Khayyat, Araek Tashkandi, Enas Jambi, Safa Habibullah, Mashael Khayyat and Kaouther Omri
Adv. Artif. Intell. Mach. Learn., 5 (4):4594-4617
1. Noura Alotaibi: Computer Science and Artificial Intelligence Department, University of Jeddah, Jeddah, 21959, Saudi Arabia
2. Manal M. Khayyat: Department of Computer Science and Artificial Intelligence, College of Computing, Umm Al-Qura University, Makkah, 24382, Saudi Arabia
3. Araek Tashkandi: Department of Information Systems and Technology, College of Computer Science and Engineering, University of Jeddah, Jeddah, 21959, Saudi Arabia
4. Enas Jambi: Computer Science and Artificial Intelligence Department, University of Jeddah, Jeddah, 21959, Saudi Arabia
5. Safa Habibullah: Department of Information Systems and Technology, Collage of Computer Science and Engineering, University of Jeddah. Jeddah, Saudi Arabia
6. Mashael Khayyat: Department of Information Systems and Technology, College of Computer Science and Engineering, University of Jeddah. Jeddah, Saudi Arabia.
7. Kaouther Omri: Department of Computer and Network Engineering, College of Computer Science and Engineering, University of Jeddah, Jeddah, 21959, Saudi Arabia
DOI: 10.54364/AAIML.2025.54255
Article History: Received on: 23-Aug-25, Accepted on: 11-Nov-25, Published on: 18-Nov-25
Corresponding Author: Noura Alotaibi
Email: nmalotaibi@uj.edu.sa
Citation: Noura M. Alotaibi, et al. Comparative Evaluation of Hybrid Multi-Classification Algorithms for Diagnosing Diabetes and Cardiovascular Diseases. Advances in Artificial Intelligence and Machine Learning. 2025;5(4):255. https://dx.doi.org/10.54364/AAIML.2025.54255
The integration of artificial intelligence (AI) into medical decision support systems (MDSS) offers transformative potential for enhancing diagnostic ac curacy, particularly in resource-constrained healthcare environments. This study presents a comparative evaluation of multi-classification algorithms to support the diagnosis of diabetes and cardiovascular diseases—two of the most prevalent and life-threatening chronic conditions worldwide. We compared the performance of six classification algorithms: support vector machine (SVM), k-nearest neighbors (KNN), linear discriminant analysis (LDA), decision tree, diagonal linear discriminant analysis (DLDA), and Naive Bayes (NB). These models were evaluated across four feature selec tion strategies: full feature set, Minimum Redundancy Maximum Relevance (MRMR), KNN-based ranking, and univariate ranking using the Chi-square test. Experimental results demonstrate that feature selection, particularly via KNN and univariate ranking, significantly improves classification accuracy. NB and DLDA achieved strong performance for diabetes prediction, reaching up to 78.7% accuracy with limited features. For cardiovascular disease prediction, the best performance of 85.6% was obtained using a tree classifier combined with univariate feature ranking. These findings highlight the effectiveness of hybrid models in medical diagnostics and underscore the importance of tailored feature selection.