Abstract
The integration of digital health technologies into diabetes management has shown the potential to improve patient outcomes by providing personalized dietary recommendations. This study aims to develop and evaluate the Diabetes-Focused Food Recommender System (DFRS), a system designed to assist individuals with diabetes in making informed food choices. Using a combination of advanced machine learning algorithms, nutrition science, and digital health technologies, DFRS generates personalized recommendations tailored to individual needs. The methodology involves data collection from diverse patient profiles and model development using Graph Neural Networks (GNN) and other machine learning techniques. Hyperparameter tuning and rigorous performance evaluation were conducted to optimize system accuracy. The results demonstrate that after optimization, GNN achieved an accuracy of 94 percent, significantly enhancing the precision of dietary recommendations. Clinical validation of the system showed a reduction in HbA1c levels, glycemic variability, and incidents of hyper- and hypoglycemia. Therefore, DFRS has proven to be an effective tool for improving dietary management in diabetes care, and its integration into clinical workflows offers the potential to enhance health outcomes and streamline healthcare delivery.