Blood Glucose Level Prediction for Diabetics Based on Nutrition and Insulin Administration Logs Using Personalized Mathematical Models

Table 3

Validation results for the full time period with various algorithmic setups as indicated (values in mmol/l).

Model training (full data test with restart)	Without training	Brute force	Fast GA	Fast GA + diurnal parameter profile
Average error	4.37	3.57	3.64	3.30
RMSE	5.72	4.74	4.82	4.43
EGA − A + B (acceptable)	78%	84%	83%	85%

The error is the difference of the CGMS reading and the model outcome.