Comparative Analysis of the Performance of Four Learning Models for Predicting Blood Glucose Levels in Type 1 Diabetic Patients Based on Blood Glucose Measurements

EasyChair Preprint no. 10766

Abstract

Type 1 diabetes requires strict blood glucose management to maintain normal levels and prevent complications. Predicting future blood glucose levels is essential for effective management. This study compares four learning models to assess their effectiveness in predicting blood glucose based on current measurements. The literature review focused on articles exploring blood glucose prediction in type 1 diabetes with minimal human intervention. The selection criteria considered the reliability and accuracy of approaches in predicting blood glucose. Performance evaluation involved measures such as root mean square error, mean absolute percentage error, coefficient of determination, relative error analysis, and sum of squares of glucose prediction errors. The datasets consisted primarily of patients with type 1 diabetes. Preliminary results showed that some approaches performed well in the short term, indicating their ability to accurately predict future blood glucose levels. Other approaches demonstrated good long-term results, highlighting the need for extended evaluation in diabetes management. Understanding the strengths and weaknesses of each approach is crucial to guide future research. This comparison underscores the importance of developing robust prediction methods tailored to clinical needs to improve the management of type 1 diabetes and prevent complications. Further research is needed to explore the most promising approaches and adapt them to the individual needs of patients.

Keyphrases: learning models, performance evaluation, Type 1 diabetes prediction

@Booklet{EasyChair:10766,
  author = {Loubna Mazgouti and Najiba El Idrissi El Amrani},
  title = {Comparative Analysis of the Performance of Four Learning Models for Predicting Blood Glucose Levels in Type 1 Diabetic Patients Based on Blood Glucose Measurements},
  howpublished = {EasyChair Preprint no. 10766},

  year = {EasyChair, 2023}}