AI-Driven Machine Learning Models for Diabetes Prediction: Emerging Trends, Techniques, and Future Prospects-A Systematic Review

Abstract

Diabetes has now become one of the chronic and widespread health problems. Therefore, this calls for better diagnostic tools, and hence we look into AI-driven machine learning models in predicting diabetes, compare various models including SVM, RF, NN, LR, and Ensemble methods. Our analysis showed that Ensemble Methods outperform the rest with 95.6% accuracy, precision at 93.5%, and recall at 94.8%, demonstrating the ability of the model in combining the advantages of more models to produce robust predictions. Neural Networks come second with an AUC-ROC score of 0.94: the stronger ones are for dealing with complex, nonlinear-type data patterns. Logistic Regression, highly interpretable, had lower performance with an accuracy of 86.7%. This paper reflects on the ability of AI to revolutionize diabetes diagnostics into being more accurate and efficient. However, these models face many challenges, especially with regards to striking a balance between accuracy and interpretability. Future work must focus on the potential real-time applicability of these models, incorporate further patient data, and even pursue hybrid approaches for even greater predictive power.