Table 2 Performance metrics for forecasting task with 95% CI.

From: A deep learning approach for blood glucose monitoring and hypoglycemia prediction in glycogen storage disease

Model

Prediction horizons

MAE (mg/dL)

RMSE (mg/dL)

MAPE (%)

TS Mixer

15 min

16.40 (13.83–18.97)

20.38 (17.8–22.96)

16.14 (13.72–18.57)

30 min

16.65 (14.32–18.99)

20.64 (18.28–22.99)

16.50 (14.03–18.97)

45 min

16.82 (14.06–19.59)

20.83 (18.08–23.58)

16.57 (13.90–19.23)

60 min

17.01 (13.55–20.46)

20.99 (17.58–24.4)

16.77 (13.59–19.95)

Patch TST

15 min

8.94 (7.91–9.97)

11.96 (10.8–13.12)

8.85 (7.79–9.9)

30 min

10.74 (9.87–11.61)

14.37 (13.26–15.47)

10.61 (9.77–11.46)

45 min

11.69 (10.76–12.62)

15.70 (14.42–16.99)

11.54 (10.67–12.41)

60 min

12.46 (11.26–13.65)

16.73 (15.11–18.35)

12.42 (11.21–13.63)

M Linear

15 min

6.01 (5.67–6.35)

8.15 (7.68–8.61)

5.95 (5.64–6.26)

30 min

9.04 (8.53–9.54)

12.32 (11.56–13.09)

8.91 (8.5–9.33)

45 min

10.38 (9.62–11.14)

14.13 (12.98–15.27)

10.23 (9.56–10.9)

60 min

11.23 (10.4–12.07)

15.19 (13.94–16.45)

11.14 (10.39–11.89)

  1. CI Confidence interval, MAE mean absolute error, RMSE root mean squared error, MAPE mean absolute percentage error.