Fig. 4
From: Machine learning glucose forecasting models for septic patients
Comparison of performance metrics for the five forecasting models: iTransformer, Crossformer, PatchTST, DLinear, and FEDformer with a lookback window length = 120 and prediction length = 15. PatchTST and DLinear achieve significantly better results, with consistently lower errors across all metrics compared to other models.
