Table 2 Comparative performance of predictive models for future glucose levels, using Root Mean Square Errors (rMSE)
Model | Dataset | rMSE-30m | rMSE-1h | rMSE-2h | |
|---|---|---|---|---|---|
LSTM | OhioT1DM | 36.022 (0.551) | 37.17 (0.513) | 38.703 (0.461) | |
RNN | 36.102 (0.552) | 37.344 (0.512) | 38.952 (0.458) | ||
GRU | 37.555 (0.554) | 38.573 (0.517) | 40.108 (0.463) | ||
Transformer | 27.886 (0.463) | 30.869 (0.462) | 36.653 (0.47) | ||
Informer | 35.197 (0.501) | 36.962 (0.485) | 40.204 (0.463) | ||
Autoformer | 36.08 (0.552) | 38.352 (0.542) | 41.395 (0.515) | ||
CGMLSM | 9.024 (0.168) | 15.895 (0.283) | 26.876 (0.43) | ||
CGM-LSM | WellDoc T1D | Internal Test | 8.403 (0.066) | 16.049 (0.118) | 28.277 (0.188) |
Temporal Test | 9.155 (0.068) | 17.013 (0.118) | 29.426 (0.184) | ||
Held-Out Test | 8.926 (0.056) | 16.905 (0.101) | 29.812 (0.16) | ||
WellDoc T2D | Internal Test | 7.441 (0.055) | 13.418 (0.094) | 22.649 (0.147) | |
Temporal Test | 8.025 (0.058) | 14.073 (0.095) | 23.216 (0.143) | ||
Held-Out Test | 7.772 (0.055) | 13.877 (0.091) | 23.494 (0.143) | ||
