GluPredKit: A Python Package for Blood Glucose Prediction and Evaluation
Managing blood glucose levels is crucial for individuals with diabetes. Historically, non-linear physiological modeling of glucose dynamics laid the groundwork for automated insulin delivery. Blood glucose prediction can be used as decision support for patients or as a component in an automated insulin delivery control strategy. Today, machine learning and deep neural networks offer new pathways for improvement, and the literature is vast on proposed models. Yet, comparing these advanced models is challenging. Differences in the datasets used for testing and how results are evaluated can make comparisons from existing studies unreliable (Jacobs et al., 2023). Additionally, many research studies do not share their code, making it hard to build upon previous work. GluPredKit addresses these issues by standardizing the pipeline steps needed for any blood glucose prediction research (see Figure 1). This includes the collection, organization, and preparation of data, as well as the ability to easily compare different models and measure their effectiveness. Additionally, the software incorporates state- of-the-art components, including the ability to integrate and standardize data from various sources, utilize existing prediction models, and apply established evaluation metrics. It also features automated generation of detailed model evaluation reports, guided by the consensus on blood glucose model evaluation (Jacobs et al., 2023).
M. K.Wolff, S. Royston, and R. Volden, "GluPredKit: A Python Package for Blood Glucose Prediction and Evaluation", published in Journal of Open Source Software, 9(101), 6904. DOI: 10.21105/joss.06904
