⬢github Python · 83 ★ +69 since we first saw it · pushed 1 d ago · MIT
0xdeadf1sh/T1DMAI
A model zoo of blood glucose forecasters trained & tested on my own data
T1DMAI is a personal research project: a model zoo of 22 blood glucose forecasting models, including a small custom transformer (with DILATE vs MSE loss variants and optional LoRA fine-tuning), trained on synthetic CGM data and the author's own type 1 diabetes data plus the public MetaboNet dataset. Honest conclusion: XGBoost wins overall, though the author's model catches hypoglycemic lows earlier and better.
Why now: It hit the Hacker News front page as a 'Show HN' post, notable for the personal angle (a diabetic building models on their own data) and the candid finding that all models perform poorly.
Who it is for: Diabetes tech researchers, patients with type 1 diabetes experimenting with CGM forecasting, and ML hobbyists interested in time-series modeling with personal data.
benchmarkbenchmarkingblood-glucoseblood-glucose-forecastingblood-glucose-predictiondeep-learningdiabetest1dmtime-seriestime-series-analysis-and-forecasting
Stars over our 583 snapshots: 14 to 83, since 6 d ago.
Where people talked about it
- Yhn Show HN: I have trained a model to predict my blood sugar 46 min ago
- Yhn Show HN: I have trained 15 blood glucose forecasting models on my own data 2 d ago
API: https://socialmediatrends-api.osmike.com/v1/repos/0xdeadf1sh/T1DMAI