An approach for Personalized Continuous Glucose Prediction with Regression Trees
Sotiris Alexiou, Ηλίας Δρίτσας, Otilia Kocsis, Κωνσταντίνος Μουστάκας, Nikos Fakotakis
Abstract
For individuals that already have been diagnosed with T2DM, accurate and timely prediction of blood glucose levels is critical to prevent hypoglycemic and hyperglycemic episodes. In clinical practice, continuous glucose monitoring (CGM) is used for the efficient management of glucose levels and, thus, diabetes. Several machine learning (ML) models can be used for the development of risk prediction models on glucose levels. Specifically, in this paper, we aim to develop an Artificial Intelligence (AI) framework for the short-term prediction of individuals' glucose values based on CGM data. The efficacy of three tree-based regressors is now under assessment in a public CGM data set. Ultimately, these models will be evaluated in real-life conditions in pilot demonstrators of the SmartWork project, whose aim is to develop an AI system for older office employees workability sustainability.