All projects
OngoingApplied AI/MLHealthcare AI · Applied Machine Learning · Explainable AI
Explainable Diabetes Risk Prediction
An explainable machine learning pipeline for diabetes risk prediction, comparing interpretable and ensemble models with threshold-based evaluation and SHAP explanations.
Objective
Build a transparent, well-evaluated risk model and understand which factors drive individual predictions.
Methods applied
- Logistic Regression
- Random Forest
- Cross-validation
- Imputation pipelines
- SHAP
Progress so far
- Dataset examination.
- Data preprocessing.
- Treatment of impossible zero values as missing.
- Imputation within cross-validation pipelines.
- Logistic Regression and Random Forest comparison.
- Five-fold cross-validation.
- Held-out evaluation.
- Threshold-based evaluation.
- SHAP analysis code and plots generated.
Next steps
- Inspect SHAP plots.
- Examine feature importance.
- Inspect individual predictions.
- Evaluate calibration.
- Review robustness.
- Complete documentation.