BLOOD GLUCOSE LEVEL PREDICTION USING GAUSSIAN PROCESS REGRESSION

Abstract

Blood glucose level prediction is an important part in the treatment of diabetes. Even though there are many different models, no data-driven models seem to provide results that are consistent enough to create a closed loop blood glucose control system. This study aims to assess the possibilities of Gaussian process regression in the prediction of blood glucose levels by the means of two experiments. In the first experiment the performance of Gaussian process regression on continuous data will be compared to Gaussian process regression on categorical data and linear regression. The second experiment analyzes the influence of the sparsity of data on gaussian process regression. The results reveal that the implemented model does not perform well compared to other research. However, the results do indicate that further research can improve the performance of Gaussian process regression on continuous data.