Prior Research
This project builds on a foundation of peer-reviewed research in machine learning, explainable AI, and diabetes prediction. Below are the key studies that informed our methodology and design decisions.
Machine Learning in Medicine
New England Journal of Medicine, 380(14), 1347–1358
Key Finding
ML models can match or exceed clinician accuracy in risk stratification tasks when trained on structured clinical data.
Relevance to our work
Justifies our choice of supervised classification over rule-based clinical screening approaches.
A Unified Approach to Interpreting Model Predictions
Advances in Neural Information Processing Systems, 30, 4765–4774
Key Finding
SHAP values provide theoretically grounded, consistent feature attribution that is superior to prior interpretability methods.
Relevance to our work
SHAP is our primary explainability method — this is the foundational paper we implement directly.
"Why Should I Trust You?": Explaining the Predictions of Any Classifier
Proceedings of the 22nd ACM SIGKDD, 1135–1144
Key Finding
LIME generates locally faithful, interpretable explanations for individual predictions of any black-box model.
Relevance to our work
LIME is our secondary explainability method used for patient-level decision interpretation.
Predicting the Future — Big Data, Machine Learning, and Clinical Medicine
New England Journal of Medicine, 375(13), 1216–1219
Key Finding
Big data and ML have the potential to transform clinical medicine by improving prediction, diagnosis, and personalised treatment.
Relevance to our work
Provides the foundational motivation for applying ML to population-level diabetes risk stratification.
Machine Learning Interpretability: A Survey on Methods and Metrics
Electronics, 8(8), 832
Key Finding
A comprehensive survey of XAI methods showing that SHAP and LIME are the most widely adopted for tabular healthcare data.
Relevance to our work
Informed our choice of SHAP and LIME over alternative XAI approaches.
Machine Learning and Data Mining Methods in Diabetes Research
Computational and Structural Biotechnology Journal, 15, 104–116
Key Finding
XGBoost and Random Forest consistently outperform other classifiers for diabetes prediction across multiple datasets.
Relevance to our work
Directly supports our selection of XGBoost as the final model for diabetes risk prediction.
IDF Diabetes Atlas, 10th Edition
International Diabetes Federation, Brussels, Belgium
Key Finding
Over 537 million adults live with diabetes globally, with Type 2 accounting for over 90% of cases. Prevalence is highest in low- and middle-income countries.
Relevance to our work
Provides the epidemiological scale and public health urgency that motivates this project.
Predicting Diabetes Mellitus with Machine Learning Techniques
Frontiers in Genetics, 9, 515
Key Finding
Ensemble methods achieve AUC > 0.85 for diabetes prediction using BRFSS-derived clinical features.
Relevance to our work
Closest prior work to our exact approach — same feature set, same problem framing, validates our methodology.