Revolutionizing Diabetes Detection: A Deep Learning Approach for Enhanced Predictive Accuracy

Abstract

Diabetes mellitus has been regarded as a very common metabolic disorder characterized by chronic hyperglycemia. It has been viewed as a major health problem and requires immediate concern with regard to its diagnosis at an early stage. This work proposes a deep learning framework that is proposed to advance the detection of diabetes, thereby proposing optimized predictive modeling using the Gated Attention Diabetes Model (GADM) with Chaos Game Optimization (CGO) for weight estimation. Our main aim in proposing this paper is a sound and robust prediction model that can estimate the onset of diabetes with high accuracy by proposing employment of advanced computational algorithms. That could enhance the model\'s ability to find some key attributes related to diabetes risk. This focused attention with feature selection not only enhances predictive accuracy but also secures interpretability of the model for use in clinical settings. Therein, the CGO optimizes the weight parameters within GADM based on chaotic dynamics for efficiently searching in the solution space for better convergence, improving thereby the classification performance. We design a classification system that will be able to effectively distinguish between diabetic and non-diabetic persons based on risk factors identified. We carry out an extensive performance analysis which involves a comparative study of the proposed model with state-of-the-art methods. Indeed, these results point to significant enhancements in predictive performance-as evidenced in sensitivity, specificity, and accuracy rates. We also conduct an extensive interpretability investigation into the same and provide insights into key characteristics and mechanisms underlying the model\'s predictions.

Introduction

The text discusses diabetes mellitus as a widespread chronic disease caused by the body’s inability to properly regulate blood glucose due to insufficient insulin production or insulin resistance. It explains the two main types—Type 1 diabetes, caused by autoimmune destruction of insulin-producing pancreatic cells, and Type 2 diabetes, which is more common and largely linked to lifestyle factors such as obesity, poor diet, and inactivity. The growing global prevalence of diabetes highlights the urgent need for early detection, accurate diagnosis, and effective management to reduce complications.

Because diabetes data is highly complex and influenced by genetic, environmental, and lifestyle factors, traditional statistical methods struggle to analyze it effectively. Machine learning (ML) and artificial intelligence (AI) offer better solutions by identifying hidden patterns in large, nonlinear datasets. These methods enable risk prediction, feature extraction, and personalized treatment planning by analyzing diverse inputs such as medical history, behavior, and biological markers.

The literature review shows extensive research on ML and deep learning for diabetes prediction, including ensemble models, neural networks, and hybrid systems. However, major challenges remain, such as imbalanced datasets, reliance on single data sources, lack of interpretability, and limited generalization across populations. Many existing models also fail to fully integrate multi-source data like lifestyle, genetic, and sensor-based information, which limits their accuracy and real-world applicability.

To address these gaps, the paper proposes a Gated Attention Diabetes Model (GADM) combined with a Chaos Game Optimization (CGO) algorithm. GADM integrates GRU-based sequential learning with attention mechanisms to better capture important features and temporal relationships in patient data. CGO is used to optimize model weights using a chaos-inspired search strategy, improving convergence, stability, and performance while avoiding local minima.

Overall, the proposed approach aims to improve diabetes prediction accuracy, handle complex medical data more effectively, and provide a more robust, optimized, and interpretable AI-based diagnostic system compared to existing methods.

Conclusion

This paper provides a unified approach for diabetes detection and classification by developing a Gated Attention Diabetes Model integrated with Chaos Game Optimization for weight estimation. This synergy between GADM and CGO resolves the issues of the conventional diabetes detection model by pushing the benchmark farther than has been done ever before regarding the accuracy and reliability of existing algorithms. These experimental results evidence that the proposed GADM-CGO framework outperforms traditional methods on several datasets, including remarkable accuracy rates and significant improvements in sensitivity and specificity metrics. Results highlight how effectively the most advanced machine learning techniques can be combined with the newest optimization strategies to solve real-world healthcare problems and, more precisely, diabetes detection and management. The results prove that the Gated Attention Diabetes Model and Chaos Game Optimization have emerged as a state-of-the-art approach to detecting diabetes, hence opening very promising directions for future research and clinical applications. Developments in this work contribute not only to an increase in predictive performance but also to ongoing efforts in exploiting artificial intelligence for improving healthcare outcomes. Further elaboration of this methodology could bring forward more advanced diagnostic techniques that may help in the early detection and management of diabetes.

References

[1] Z. Zhang, K. A. Ahmed, M. R. Hasan, T. Gedeon, and M. Z. Hossain, “A Deep Learning Approach to Diabetes Diagnosis,” arXiv preprint arXiv:2403.07483, 2024. [2] X. Zou, Y. Luo, Q. Huang, Z. Zhu, Y. Li, X. Zhang, X. Zhou, and L. Ji, “Differential effect of interventions in patients with prediabetes stratified by a machine learning?based diabetes progression prediction model,” Diabetes, Obesity and Metabolism, vol. 26, no. 1, pp. 97-107, 2024. [3] K. Dharmarajan, B. Sharma, S. Chowdhury, and I. B. Dhaou, \"Diabetes Prediction using Enhanced and Optimized DRL-CNN Approaches.\" pp. 278-283. [4] P. Ghadekar, Y. Munde, N. Choudhary, M. Sayyed, A. Vidhale, and N. Zanwar, “Early Diabetes Detection and its Effect on Heart using Random Forest,” Grenze International Journal of Engineering & Technology (GIJET), vol. 10, 2024. [5] A. Khanam, and F. S. Masoodi, \"Early Detection of Type-2 Diabetes Mellitus using Machine Learning based Prediction Models.\" pp. 1398-1403. [6] Z. Mushtaq, M. F. Ramzan, S. Ali, S. Baseer, A. Samad, and M. Husnain, “Research Article Voting Classification-Based Diabetes Mellitus Prediction Using Hypertuned Machine-Learning Techniques,” 2022. [7] M. S. Reza, U. Hafsha, R. Amin, R. Yasmin, and S. Ruhi, “Improving SVM performance for type II diabetes prediction with an improved non-linear kernel: Insights from the PIMA dataset,” Computer Methods and Programs in Biomedicine Update, vol. 4, pp. 100118, 2023. [8] P. Verma, and A. Khatoon, \"Data Mining Applications in Healthcare: A Comparative Analysis of Classification Techniques for Diabetes Diagnosis Using the PIMA Indian Diabetes Dataset.\" pp. 1-5. [9] I. Wajahat, F. Keshtkar, and S. A. C. Bukhari, \"Advancing Precision Healthcare Analytics: Machine Learning Approach-es for Diabetes Prognosis using the PIMA Indian Dataset.\" [10] L. Xie, “Pima Indian Diabetes Database and Machine Learning Models for Diabetes Prediction,” Highlights in Science, Engineering and Technology, vol. 88, pp. 97-103, 2024. [11] D. Baswaraj, C. V. N. Raju, P. C. S. Reddy, A. Kiran, M. K. Shaik, and D. A. Kumar, \"An Efficient Proposal for Deep Learning-Based Diabetes Prediction.\" pp. 1-6. [12] L. Al-Dabbas, and A. A. Abu-Shareha, “Early Detection of Female Type-2 Diabetes using Machine Learning and Oversampling Techniques,” Journal of Applied Data Sciences, vol. 5, no. 3, pp. 1237-1245, 2024. [13] S. Alketbi, A. Albadawi, and Z. Al Aghbari, \"A Deep Learning-Based Ensemble Method for Early Detection of Diabetes.\" pp. 01-06. [14] A. C. Kulkarni, and B. V. Satyasrikar, \"Effects of Pre-Processing Techniques on Diabetes Detection Using Machine Learning.\" pp. 1-4. [15] A. A. Abu-Shareha, H. Qutaishat, and A. Al-Khayat, “A Framework for Diabetes Detection Using Machine Learning and Data Preprocessing,” Journal of Applied Data Sciences, vol. 5, no. 4, pp. 1654-1667, 2024. [16] H. Naz, R. Nijhawan, and N. J. Ahuja, \"DT-DL Based Hybrid Approach for Early Detection of Diabetes Using PIMA Dataset.\" pp. 1-6. [17] S. Larabi-Marie-Sainte, L. Aburahmah, R. Almohaini, and T. Saba, “Current techniques for diabetes prediction: review and case study,” Applied Sciences, vol. 9, no. 21, pp. 4604, 2019. [18] M. T. García-Ordás, C. Benavides, J. A. Benítez-Andrades, H. Alaiz-Moretón, and I. García-Rodríguez, “Diabetes detection using deep learning techniques with oversampling and feature augmentation,” Computer Methods and Programs in Biomedicine, vol. 202, pp. 105968, 2021. [19] M. K. Hasan, M. A. Alam, D. Das, E. Hossain, and M. Hasan, “Diabetes prediction using ensembling of different machine learning classifiers,” IEEE Access, vol. 8, pp. 76516-76531, 2020. [20] M. H. Fakhar, M. Z. Baig, A. Ali, M. T. A. Rana, H. Khan, W. Afzal, H. U. Farooq, and S. Albouq, “A Deep Learning-based Architecture for Diabetes Detection, Prediction, and Classification,” Engineering, Technology & Applied Science Research, vol. 14, no. 5, pp. 17501-17506, 2024. [21] B. Elaziz, Y. Ommane, M. Eddabbah, and Y. Laaziz, \"Enhancing Diabetes Detection Through Data Preprocessing: A Comparative Analysis of Machine Learning Algorithms.\" pp. 1-6. [22] A. Abdalrada, A. F. Neamah, and H. Murad, “Predicting Diabetes Disease Occurrence Using Logistic Regression: An Early Detection Approach,” Iraqi Journal For Computer Science and Mathematics, vol. 5, no. 1, pp. 160-167, 2024. [23] Q. W. Khan, K. Iqbal, R. Ahmad, A. Rizwan, A. N. Khan, and D. Kim, “An intelligent diabetes classification and perception framework based on ensemble and deep learning method,” PeerJ Computer Science, vol. 10, pp. e1914, 2024. [24] A. A. Alhussan, D. S. Khafaga, E.-S. M. El-kenawy, M. M. Eid, and A. A. Abdelhamid, “Hybrid waterwheel plant and stochastic fractal search optimization for robust diabetes classification,” AIP Advances, vol. 14, no. 6, 2024. [25] L. Z. Chee, S. Sivakumar, K. H. Lim, and A. A. Gopalai, “Gait acceleration-based diabetes detection using hybrid deep learning,” Biomedical Signal Processing and Control, vol. 92, pp. 105998, 2024.

Copyright

Copyright © 2026 Dr. M. Jayakumar, Dr. P. Jagatheesh. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.