Research on Blood Glucose Concentration Prediction Method Based on CEEMDAN-PSO-LSTM

Blood glucose prediction plays a key role in the automated treatment of diabetic patients. In order to improve the prediction accuracy of blood glucose concentration, this paper proposes a short-term blood glucose concentration prediction model that combines the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) algorithm with a Particle Swarm Optimization Long Short-Term Memory (PSO-LSTM) neural network. Firstly, the CEEMDAN algorithm is used to decompose the blood glucose concentration time series into sub-sequence components. Then, each sub-sequence component is predicted using the PSO-LSTM neural network separately. Finally, the predicted results of each Intrinsic Mode Function (IMF) component are accumulated to obtain the final prediction value of the blood glucose concentration. Experimental analysis shows that the proposed model achieves the best performance in terms of the four prediction evaluation indicators (R2, RMSE, MAE, SMAPE), validating the superiority of the CEEMDAN-PSO-LSTM model.