NARMA modeling of subcutaneous tissue glucose time-series using neural networks

Absfract A radial basis function (RBF) neural network for modeling glucose dynamics with a non-linear autoregressive moving avenge model (NARMAI is presented. Subcutaneous tissue glucose from healthy bubjects was frequently sampled using portable device for open tissue perfusion and the time-series were used for identification of the NARMA model. Validity tests for non-linear models applied on a test data set demonstrate that a parsimoneous RBF network can be obtained for modeling subcutaneous tissue glucose dynamics.