Abstract
In recent years, diabetes has become an important problem for humanity. Diabetes is a chronic disease and affects around 500 million people worldwide. People with diabetes who are unrealize of this live with the disease until they see time, it can also damage other organs. Therefore, diabetes has required lifelong following carefully. Blood glucose (BG) control is widely known as one of the diabetes management ways. With continuous glucose management (CGM) systems, it has become easier to monitor blood sugar. These systems record a blood glucose measurement every 5 minutes. Thus, it allows us to build smart healthcare systems learnt from the data. Deep learning approaches has recently been applied in healthcare researchs. Blood glucose prediction is a time series problem. Deep neural networks that learn from data are used to solve this problem. In this study, three different neural networks; Recurrent Neural Networks (RNN), Gated Recurrent Units (GRU) and Long Short-Term Memory (LSTM) based models are compared on OhioT1DM dataset that consist of 12 patients. Blood glucose forecast horizon is decided as 30 minutes. Methods have evaluated using the Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) metrics.
Keywords
Subject Areas
Citations by Year
Related YTU Activities
Sustainability activities and projects connected to this article.
