BMC Medical Informatics and Decision Making

Table 4 The table represents the prediction comparison of proposed models over 30 and 60 min of the prediction horizon

From: Stacked LSTM based deep recurrent neural network with kalman smoothing for blood glucose prediction

Patient ID	PH = 30 min		PH = 60 min
Patient ID	Single LSTM	Stacked LSTM	Single LSTM	Stacked LSTM
# 559	18.03	17.85	31.89	31.55
# 563	19.20	18.65	31.01	30.42
# 570	16.63	15.94	26.28	25.74
# 575	21.12	20.93	32.90	31.97
# 588	18.07	17.71	31.11	30.45
# 591	20.71	20.35	32.06	31.80
Mean (RMSE)	18.96	18.57	30.88	30.32

Here, the models are trained with the raw CGM readings along with the other three features mentioned before. The first one is the RNN model having a single LSTM layer, whereas the second one is stacked LSTM based deep RNN model

Back to article page

ISSN: 1472-6947

Contact us

General enquiries: journalsubmissions@springernature.com