Glucose-tracking: A postprandial glucose prediction system for diabetic self-management

Up to today, there are around 400 million diabetics in the world. In China, there are more than 100 million diabetics. How to help them track and manage real-time glucose level is significant to control diabetic progression. As well known, the glucose level is directly related with food, while the conventional tracking glucose is depending invasive or minimally invasive methods. This phenomenon causes many problems to diabetics for food selection and glucose monitoring. To solve this challenge, we are motivated to propose a postprandial glucose prediction model, which can connect food selection and diabetic self-management seamlessly, without involving invasive glucose monitoring. In this paper, we first build up a glucose tracking system based on Android platform. And then, we conceive a postprandial glucose prediction model based on machine learning methods, in which sample data from diabetics' diet are collected and analyzed for predicting a postprandial glucose level of a user. To verify our model, we compared the prediction results with our clinical tests by using traditional glucose monitoring methods. In order to extract a reliable glucose prediction model, two kinds of linear regression algorithms are adopted, which confidence margin is bounded within 20% that matching with the FDA standard. Our experiment results show that this prediction-based glucose tracking system is helpful to diabetics, and can be used as an auxiliary tool to control their glucose and diet. ? 2015 IEEE.EI9-1