Deep Learning for Blood Glucose Prediction: CNN vs LSTM

El Idrissi, Touria; Idri, Ali

To manage their disease, diabetic patients need to control the blood glucose level (BGL) by monitoring it and predicting its future values. This allows to avoid high or low BGL by taking recommended actions in advance. In this study, we propose a Convolutional Neural Network (CNN) for BGL prediction. This CNN is compared with Long-short-term memory (LSTM) model for both one-step and multi-steps prediction. The objectives of this work are: 1) Determining the best configuration of the proposed CNN, 2) Determining the best strategy of multi-steps forecasting (MSF) using the obtained CNN for a prediction horizon of 30 min, and 3) Comparing the CNN and LSTM models for one-step and multi-steps prediction. Toward the first objective, we conducted series of experiments through parameter selection. Then five MSF strategies are developed for the CNN to reach the second objective. Finally, for the third objective, comparisons between CNN and LSTM models are conducted and assessed by the Wilcoxon statistical test. All the experiments were conducted using 10 patients’ datasets and the performance is evaluated through the Root Mean Square Error. The results show that the proposed CNN outperformed significantly the LSTM model for both one-step and multi-steps prediction and no MSF strategy outperforms the others for CNN.