Glucose prediction for type 1 diabetes using KLMS algorithm

For the patients with type 1 diabetes (T1D), it is very important to keep their blood glucose concentration in the normal level by insulin injections. As the glucose level can be checked consistently by continuous glucose monitoring (CGM) system, it enables estimation of near-future glucose prediction by developing a reliable prediction model. In this paper, a kernel-based adaptive filtering algorithm is applied to build prediction models for glucose prediction. Then, the details for selecting a proper kernel function are also investigated. Finally, the efficiency of the proposed kernel-based forecasting method is evaluated in the short-term blood glucose prediction. The relative results are also analyzed to outline the performance of the proposed KLMS algorithm.