Predictive Modeling of Blood Glucose Levels in Type 1 Diabetes Patients Using Ensemble Learning Algorithms with Multisource Fusion Data

Type 1 diabetes is a chronic autoimmune disorder characterized by the destruction of insulin-producing beta cells in the pancreas, leading to an absolute deficiency of insulin. This deficiency impairs the body's ability to regulate blood glucose levels, resulting in hyperglycemia. Without meticulous management, including exogenous insulin administration, the condition can result in severe acute and chronic complications, posing significant risks to health and survival. Accurate prediction of blood glucose (BG) levels is essential for managing Type 1 Diabetes (T1D). This study aims to develop an effective algorithm that can predict BG levels one hour in advance using participants' data from the past six hours. We utilize a comprehensive dataset collected from young adults in the UK, including continuous glucose monitoring (CGM), insulin pump usage, carbohydrate intake, and activity data from smartwatches. Our model integrates these multisource physiological parameters to provide timely and accurate predictions, which can help patients better manage their condition. The performance of our model is evaluated using the Root Mean Square Error (RMSE) metric, demonstrating significant accuracy in BG level predictions.