Obtaining Difference Equations for~Glucose Prediction by~Structured Grammatical Evolution and~Sparse Identification

Abstract

Diabetes is one of the most common and difficult non-communicable diseases to deal with in our days. People with diabetes need to keep their glucose levels within a certain range to avoid health complications. Some patients must inject insulin to regulate their glucose levels, and estimating the necessary dose is not an easy task. In this paper, we investigate how to obtain expressions that predict glucose levels using variables such as previous glucose values, food ingestion (in carbohydrates), basal insulin dosing, and dosing of bolus of insulin. This paper proposes the combination of structured grammatical evolution and sparse identification to obtain difference equations governing the dynamics of the glucose levels over time. Glucose prediction serves as a tool for deciding the most convenient insulin dosing. Our technique produces promising results that provide explainable equations and use information usually managed by people with diabetes.

J. Manuel Colmenar

Full Professor

My research interests are focused on metaheuristics applied to optimization problems. I have worked on different combinatorial optimization problems applying trajectorial algorithms such us GRASP or VNS. Besides, I am very interested in applications of Grammatical Evolution, specifically in model and prediction domain, as alternative to machine learning approaches.