Multi-Parametric Glucose Prediction Using Multi-Layer LSTM

Authors : Ömer Atılım Koca, Volkan Kiliç

Pages : 169-175

View : 54 | Download : 70

Publication Date : 2023-12-15

Article Type : Research Paper

Abstract :Diabetes causes irregular glucose levels, such as hyperglycemia (high glucose) and hypoglycemia (low glucose), which affect the quality of life of diabetes patients. Early detection of hyperglycemia and hypoglycemia is important for effective management of the disease. In recent years, progress has been made in the development of artificial intelligence-based tools for effective diabetes management. These tools aim to predict glucose levels before they reach critical levels, enabling people with diabetes to take proactive measures to keep their glucose levels within a healthy range. However, most of these tools use single-layer architectures and rely only on glucose measurement as a predictive parameter, thus resulting in low predictive accuracy. Here, this paper proposes a multi-layer Long-Short Term Memory (LSTM)-based model for glucose prediction. The proposed model was tested on the OhioT1DM dataset and the lowest Root Mean Square Error value was obtained as 14.364 mg/dL for glucose prediction over a 30-min prediction horizon. The results demonstrate the performance of the proposed system, which uses a multi-layer LSTM algorithm to overcome the complex memory operations associated with multi-parameter prediction.
Keywords : Yapay Zeka, LSTM, Diyabet, Glikoz Tahmini