Ensemble Regression for Blood Glucose Prediction

Background: Predicting blood glucose present commonly many challenges when the designed models are tested under different contexts. Ensemble methods are a set of learning algorithms that have been successfully used in many medical fields to improve the prediction accuracy. This paper aims to review the typology of ensembles used in literature to predict blood glucose. Methods: 32 papers published between 2000 and 2020 in 6 digital libraries were selected and reviewed with regard to: years and publication sources, integrated factors and data sources used to collect the data and types of ensembles. Results: This review results found that this research topic is still recent but is gaining a growing interest in the last years. Ensemble models used often blood glucose, insulin, diet and exercise as input to predict blood glucose. Moreover, both homogeneous and heterogeneous ensembles have been investigated. Conclusions: An increasing interest have been devoted to blood glucose prediction using ensemble methods during the last decade. Several gaps regarding the design of the reviewed ensembles and the data collection process have been identified and recommendations have been formulated in this direction. © 2021, The Author(s), under exclusive license to Springer Nature Switzerland AG.