GlucoTune: A Unified Framework for Blood Glucose Preprocessing, Forecasting, and Benchmarking in Diabetes

GlucoTune introduces an extensible framework for reproducible blood-glucose time-series research, covering preprocessing, model training, evaluation, and benchmarking. Its preprocessing pipelines are specified through portable YAML configuration files, allowing experiments to be reconstructed from original datasets without redistributing sensitive preprocessed data. The framework offers standardized wrappers for public datasets, interfaces for blood-glucose and general time-series forecasting models, and a leaderboard comparing datasets, preprocessing configurations, and forecasting methods. The authors report comprehensive experiments and a usability user study, but the abstract does not provide quantitative results.