Using Gaussian Processes to Monitor Diabetes Development

This paper uses Gaussian process techniques to model time series data of HbA1c level, a common measure to monitor or screen diabetes. The HbA1c level estimates how well blood sugar is under control. To facilitate the control of diabetes, we develop a patient-level model to individually predict the development of the disease for each patient. Gaussian processes represent a successful machine learning technique known for their flexible modeling abilities and high predictive performances. This approach allows multi-dimensional inputs and assigns a confidence score to the predictions, accounting for temporal uncertainty of time series data. The purpose of this paper is to discuss the use of the Gaussian process technique, previously unseen in diabetes research, to monitor the development of the disease.