Blood Glucose Prediction for Type 1 Diabetes using Machine Learning
In this thesis, walk forward testing was used to evaluate the performance of two long short-term memory (LSTM) models for predicting blood glucose values for patients with type 1 diabetes. The models were compared with a support vector regression (SVR) model as well as with an auto regressive integrated moving average (ARIMA) model, both of which have been used in related research within the area. The best performing long short-term model produces results similar to those of the SVR model and it outperforms the ARIMA model for all prediction horizons. In contrast to models in related research, our LSTM model is trained to assign a level of confidence to each prediction, adding an edge in practical usability.
Christian Meijner, Simon Persson