Evaluating artificial intelligence approaches to forecast and interpret continuous glucose monitoring data in type 1 diabetics

People with type 1 diabetes often experience rapid changes in blood sugar, especially during and after exercise. These changes can be difficult to predict, even with continuous glucose monitors (CGMs) and modern diabetes technology. Our project aims to develop and test a computer model that can help forecast short-term blood sugar changes using information people already collect—such as their CGM readings, heart rate, and details about their physical activity.

We will use the Type 1 Diabetes Exercise Initiative (T1DEXI) dataset, which contains anonymized information from adults with type 1 diabetes who wore CGMs and activity trackers while going about their daily lives. By analyzing these real-world data, we will test how well our model predicts blood sugar during different types of exercise and daily activities. We will also examine which factors—such as exercise intensity, duration, meals, or stress—most strongly influence changes in blood sugar.

The goal of this research is to improve short-term glucose forecasting in ways that could eventually support safer exercise planning and better everyday diabetes management. This study does not provide medical advice or make individual recommendations; it is solely a scientific analysis of anonymized data.