BMC Medical Informatics and Decision Making

Table 2 BG prediction experimental performance in terms of RMSE, with Stacked LSTM model for different feature set as input

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

Patient ID	G	G, C	G, C, I	G, C, I, S	G, C, I, B	G, C, I, SL	G, C, I, GSR	G, C, I, HR
# 559	19.42	18.73	18.03	17.85	18.07	19.04	19.01	18.87
# 563	19.07	18.92	18.76	18.65	18.74	19.03	19.12	19.09
# 570	16.26	16.11	16.12	15.94	16.13	16.17	16.86	16.23
# 575	22.68	21.89	21.02	20.93	21.08	21.66	22.73	21.58
# 588	19.12	18.64	18.19	17.71	18.17	18.21	19.24	18.77
# 591	23.41	21.87	20.39	20.35	20.42	20.38	22.47	21.42
Mean RMSE	19.99	19.36	18.75	18.57	18.77	19.08	19.90	19.33

G CGM value, C carbohydrate info, I insulin from Bolus, S step info, B insulin from basal, SL sleep info, GSR galvanic skin response, HR heart rate
Bold value represents the best prediction result among experiments with different feature set

Back to article page

ISSN: 1472-6947

Contact us

General enquiries: journalsubmissions@springernature.com