GlucoWise: code and trained models for multimodal ML prediction of postprandial glycemic peaks in adults with diabetes

Reproducibility package for the Journal of Diabetes Science and Technology (SAGE Publishing / Diabetes Technology Society) submission "Multimodal Machine Learning Prediction of Postprandial Glycemic Peaks in Adults With Diabetes: The GlucoWise Model". Open-access publication funded through the Sage 90% Transformative Agreement (Consorcio Colombia 2026) via Corporación Universitaria Minuto de Dios (Uniminuto). Contains the complete training notebooks, the trained calibrated HistGradientBoosting model and cluster-aware experts, isotonic calibrators, K-Means centroids (K=4 glycemic phenotypes), the processed analytical CSV (event-aligned feature matrix with the binary outcome is_high_peak_120 for the BIG IDEAs Lab v1.1.2 cohort; 6,268 events from 15 adults with type 1 or type 2 diabetes), the figure-generation scripts (LOPO + bootstrap + calibration + DCA + SHAP), the manuscript LaTeX source, and the TRIPOD+AI checklist. All random seeds fixed at 42. The raw BIG IDEAs Lab v1.1.2 data are not redistributed and remain available at their canonical source on PhysioNet (DOI 10.13026/MWGF-H309). Code released under the MIT licence; derived data under CC-BY 4.0.