Blood Glucose Prediction Based on ARIMA Time-Series Machine Learning Model
Keywords:ARIMA, Time-series Blood Glucose Prediction, Cobelli Model, Type 1 Diabetes Melitus
Glucose levels prediction is a difficult task commonly faced by people with diabetes, a chronic health condition that affects how a human body synthesizes food. The glucose levels in the human body depend on a variety of factors, so the patient always assumes the risk of making incorrect calculations. Nowadays, using new technologies such as Artificial Intelligence (AI) or Machine Learning (ML), these calculations can be supported and eased by the application of prediction systems. Time series modelling involves developing models used to describe the observed time series and understand the "why" behind its dataset. In recent years, there has been a growing trend of applying machine learning algorithms to time-series predictions. Machine learning approaches have been applied to the prediction of blood glucose levels in several studies. However, it is hard to compare the performance of different prediction approaches used in these references, either classical regression or machine learning based models, since different datasets were used by different studies. In this work, through the simulation of an open-loop insulin delivery system based on Cobelli type 1 diabetic model, time-series blood glucose datasets were generated and were used to train auto regressive integrated moving average (ARIMA) ML model for the blood glucose prediction. This work is believed could give more insight in improving the diabetes management and treatment. The study's end goal is to examine the outcomes and performance of the constructed machine learning-based system. The performance of the machine learning model was evaluated through the Mean Squared Error (MSE), Mean Average Error (MAE) and Root Mean Squared Error (RMSE). It was found that the mean errors of MSE = 643, MAE = 19.83 and RMSE = 25.04 for 70:30 of train and test data splitting were lower than the other two ratios of 80:20 and 90:10 after the seasonality removal was conducted.