Project 3: Knowledge distillation across datasets for blood glucose forecasting
Won Blood Glucose Forecasting Challenge (BGLP) to accurately predict glucose levels of patients in OhioT1DM dataset. Our model ranked 2nd among 15 teams from all over the world with root mean square error under 20 units. Work was presented at 24th European Conference on Artificial Intelligence, Knowledge Discovery in Healthcare.
Competition details: Knowledge Discovery in Healthcare Data, 2020