Genetic Algorithm Optimized MLP Model for Diabetes Prediction

This study proposes an optimized and interpretable framework for diabetes prediction by combining feature selection, a genetic algorithm-optimized Multilayer Perceptron, and explainable AI techniques. Feature selection helped to reduce data dimensionality while identifying the most informative clinical variables, which made the subsequent model training more focused and efficient. Using a genetic algorithm to tune hyperparameters appeared to improve the MLP’s predictive performance noticeably, while explainable AI methods such as SHAP and LIME offered valuable insights into how individual features influenced the predictions, adding a layer of transparency that is often missing in complex models.      Experimental results suggest that the GA-MLP framework consistently outperforms conventional MLPs across metrics like accuracy, precision, recall, and F1-score. This indicates that the approach is not only robust but also potentially reliable for real-world healthcare applications, though some caution is warranted when generalizing beyond the current dataset. The framework thus presents a practical, data-driven tool that could assist healthcare professionals in early diabetes detection, while also allowing them to understand why the model reaches its conclusions.      Looking forward, there are several directions that could further strengthen this work. Incorporating larger, multi-center, or longitudinal datasets would likely improve generalization and ensure the model performs well across diverse populations. Exploring hybrid approaches, for example combining GA-MLP with ensemble methods or deeper architectures such as CNNs or LSTMs, could help capture more complex patterns in healthcare data. Integrating multi-modal information, including genetic, imaging, and lifestyle factors, may further enhance prediction accuracy and allow for more personalized recommendations. Additionally, developing real-time clinical decision support systems that leverage both the predictions and their explainable outputs could enable continuous patient monitoring. Finally, investigating advanced XAI techniques could reveal deeper insights into feature interactions and model reasoning, potentially increasing clinician trust and adoption in practical healthcare settings.

[1]     Almutairi, E., Abbod, M., & Hunaiti, Z. (2025). Prediction of Diabetes Using Statistical and Machine Learning Modelling Techniques. Algorithms, 18(3), 145. [2]     Qteat, H., & Awad, M. (2021). Using Hybrid Model of Particle Swarm Optimization and Multi-Layer Perceptron Neural Networks for Classification of Diabetes. International Journal of Intelligent Engineering & Systems, 14(3). [3]     ElSayed, N. A., Aleppo, G., Bannuru, R. R., Bruemmer, D., Collins, B. S., Ekhlaspour, L., ... & American Diabetes Association Professional Practice Committee. (2024). 2. Diagnosis and classification of diabetes: Standards of care in diabetes—2024. Diabetes Care, 47. [4]     Ogurtsova, K., Guariguata, L., Barengo, N. C., Ruiz, P. L. D., Sacre, J. W., Karuranga, S., ... & Magliano, D. J. (2022). IDF diabetes Atlas: Global estimates of undiagnosed diabetes in adults for 2021. Diabetes research and clinical practice, 183, 109118. [5]     Dayal, K., Shukla, M., & Mahapatra, S. (2023). Disease prediction using a modified multi-layer perceptron algorithm in diabetes. EAI Endorsed Trans Pervasive Health Technol, 9. [6]     Nodoust, A., & Ghatari, A. R. (2024). Assessment of a Hybrid Machine Learning Algorithm in Healthcare Management for Predicting Diabetes Disease. International Journal of Supply and Operations Management, 11(4), 462-482. [7]     El-Kenawy, E. S. M., Alhussan, A. A., Khafaga, D. S., Eid, M. M., & Abdelhamid, A. A. (2024). Optimized classification of diabetes using dynamic waterwheel plant optimization algorithm. Scientific Reports, 14(1), 23386. [8]     Dweekat, O. Y., & Lam, S. S. (2023). Optimized design of hybrid genetic algorithm with multilayer perceptron to predict patients with diabetes. Soft Computing-A Fusion of Foundations, Methodologies & Applications, 27(10). [9]     Deberneh, H. M., & Kim, I. (2021). Prediction of type 2 diabetes based on machine learning algorithm. International journal of environmental research and public health, 18(6), 3317. [10]   Chushig-Muzo, D., Calero-Díaz, H., Lara-Abelenda, F. J., Gómez-Martínez, V., Granja, C., & Soguero-Ruiz, C. (2024). Interpretable data-driven approach based on feature selection methods and GAN-based models for cardiovascular risk prediction in diabetic patients. IEEE Access, 12, 84292-84305.