A Brief Survey on Machine Learning Techniques for Blood Glucose Prediction

Abstract

Prognosis of blood glucose level is one of the significant roles in the effective treatment of diabetes which is a chronic metabolic disease that affects a good number of people in the world. Machine learning methods were of great interest in the recent as it is able to detect and predict the glucose level effectively based on the large range of features such as historical records, physiological indicators, and lifestyle data. It is a survey study that provides a detailed insight into the current investigation on blood glucose level forecast using machine learning methods. And the necessity of the accuracy of the prediction of blood glucose level in the diabetes care and the restrictions of the current methods. Roughly half of the machine learning methods applied to this field, similar support vector machines, random forests, regression models, decision trees and neural networks. Deep learning models are remembered with specific emphasis due to promising applications in the ability to learn time dependence and the ability to follow complex patterns in blood glucose records. Recurrent neural networks (RNNs), convolutional neural networks (CNNs), and long short-term memory (LSTM) networks are particular examples of deep learning models. The questionnaire proceeds with the recapping of the current issues as well as unresolved research inquiries in the field of machine learning-based blood glucose level detection and prediction.