Towards Transparent and Accurate Diabetes Prediction Using Machine Learning and Explainable Artificial Intelligence

Machine learning can accurately predict diabetes risk: The study achieved 92.50% test accuracy and an ROC-AUC score of 0.975, demonstrating that AI models can reliably identify individuals at risk of diabetes mellitus when properly trained and validated.

Explainability is essential for healthcare AI: By combining machine learning with explainable AI (XAI) tools, researchers could identify which factors the model actually used to make predictions—BMI, age, general health, income, and physical activity—making the AI's decision-making process transparent rather than a "black box."

Lifestyle and socioeconomic factors matter for diabetes risk: The study identified that modifiable factors like physical activity and income level, alongside clinical measures like BMI and age, were among the top predictors, suggesting these should be considered in prevention and management strategies.

Data quality techniques improve AI reliability in healthcare: The use of SMOTE (Synthetic Minority Oversampling Technique) to handle imbalanced data demonstrates that careful data preprocessing is crucial for building trustworthy AI systems that don't favor common cases over rare but important ones.

Transparent AI could improve clinical adoption: By making machine learning models interpretable and accurate, this framework suggests AI tools could be more readily trusted and integrated into real healthcare systems, where clinicians need to understand why a prediction is being made.