Methods to predict blood glucose level for type 2 diabetes patients

Diabetes has turned into an ever-expanding global health concern for people's health and well-being. Though it is almost incurable but proper management can contribute toward keeping it under-control to allow affected people to lead a healthy life for decades. To manage diabetes in a smarter way, accurate and hassle free prediction of blood glucose level (BGL) is paramount important for patients. Toward this goal, we propose three prediction models (Linear Regression Model, SVR Model, Weibull Distribution model) each of which uses the current and previous days BGLs to predict the next day's fasting BGL. The uniqueness of our models are that they require only a single type of data (BGLs) with minimum numbers of data inputs to make the prediction without compromising the accuracy level. Among our three models, SVR model performs better with average 3.24 mmol/L RMSE. We are considering to predict BGLs specifically for Type 2 diabetes patients as the Type 2-diabetes-affected people constitute major segment among the diabetes victims and it has so far not received the level of attention of the health workers and researchers.