An adaptive glucose prediction method using auto-regressive (AR) model and Kalman filter

Glucose prediction is a clinically important task for managing an artificial pancreas. In the literature, some methods have been proposed for this task, where an auto-regressive (AR) model is wide considered a promising structure for the prediction. However, the online identification of the parameters for the AR model remains an open issue. In this manuscript, a Kalman filer (KF) was implemented to identify the parameters for the AR model online, and this novel combination is in fact an adaptive glucose prediction algorithm. The proposed method are compared with a standard Kalman predictor on clinical data, and the experiment results demonstrate that the proposed method has superior prediction performance than the traditional methods.