Skip to main content

Table 4 Mean Absolute Error (MAE) on the test set. Results are reported as median [25th-75th] percentile. Abbreviations: FP/day, false positives per day; TG, time gain; rAR, recursive Autoregressive model; ARIMA, Autoregressive Integrated Moving Average; NN, Neural Network; LSTM, Long Short-Term Memory Neural Network; CNN-LSTM, Convolutional Long Short-Term Memory Neural Network; RF, Random Forest; LGB, LightGBM

From: Towards a decision support system for post bariatric hypoglycaemia: development of forecasting algorithms in unrestricted daily-life conditions

 

MAE [mg/dL]

Model

PH = 15 min

PH = 20 min

PH = 25 min

PH = 30 min

rAR

13.18 [11.46–16.10]

18.18 [15.97–22.70]

23.62 [20.87–28.76]

29.77 [24.99–36.82]

ARIMA

11.37 [10.31–13.25]

14.59 [13.00–17.40]

17.54 [15.36–21.05]

19.59 [17.29–23.49]

NN

10.63 [9.70–11.87]

13.63 [12.48–15.02]

16.09 [14.81–18.19]

18.10 [16.36–20.74]

LSTM

10.22 [9.47–11.56]

13.35 [12.35–14.59]

15.89 [14.46–17.99]

18.18 [16.19–20.49]

CNN-LSTM

10.28 [9.49–11.59]

13.35 [12.25–14.68]

15.75 [14.41–17.83]

18.01 [16.07–20.59]

RF

10.85 [9.86–12.39]

13.83 [12.58–16.21]

16.81 [15.59–19.51]

19.16 [17.44–21.87]

LGB

10.90 [9.97–12.46]

14.01 [12.50–16.15]

16.87 [15.41–19.08]

18.88 [17.12–21.55]