ADS Capstone Chronicles Revised

4

ultimately having to correct glucose levels with the required insulin 2.1 Problem Identification and Motivation The goal to bridge the information gap for those living with diabetes is motivated by the ability to improve overall health and quality of life. According to the National Institute of Diabetes and Digestive and Kidney Diseases, around 8.7 million adults, ages 18 years or older, were discovered to have had undiagnosed diabetes (U.S. Department of Health and Human Services, n.d.). Not only can this predictive modeling recommendation system help those aware of their diabetic condition, but also those who may not be. Additionally, the presence of diabetes is associated with significant financial and economic impacts. The “average medical expenditures among people with diagnosed diabetes were 2.6 times higher than what expenditures would be in the absence of diabetes” (American Diabetes Association, 2024, para. 5). A resource such as this recommendation system not only assists people with diabetes in health management, but can also help people who are prediabetic or individuals who simply want to adopt a healthier lifestyle. 2.2 Definition of Objectives To assist the millions of individuals who are diabetic or prediabetic, the team is proposing a predictive modeling recommendation system that implements aspects of classification modeling. This recommendation system will label certain meals, or ingredients, as safe depending on the measure of harmful nutritional elements in relation to one's glucose levels. With the nature of the recommendation system, accuracy must be optimized to deem the data science project a success. To properly optimize

accuracy metrics such as precision and recall, a comprehensive list of features will be developed, classification thresholds will be monitored, and hyperparameters will be tuned as needed. 3 LiteratureReview When designing a personalized food recommendation system for diabetic patients, it is vital to align our model with current research on nutritional management for diabetes and recommendation systems in healthcare. This literature review explores the methodical process of existing studies on dietary guidelines for diabetes, systemic reviews on current processes for individuals to obtain nutritional information, the application of machine learning in dietary recommendations, data integration methodologies for health-related recommendation systems, and further advances within machine learning and diabetes predictions. These sources provide a foundation for building a system that not only adheres to established nutritional guidelines but also effectively combines personalized health data with real-time dietary recommendations. 3.1 Dietary Advice for Individuals with Diabetes To ensure the accuracy and health benefits of our food recommendation system, it is important to base our model on established dietary guidelines for diabetic patients. Dietary Advice for Individuals with Diabetes (Reynolds & Mitri, 2024) has an in-depth examination of evidence-based nutritional recommendations that are tailored for individuals who are managing their diabetes. This guide, which is structured for healthcare providers, puts emphasis on dietary strategies for patients with diabetes that align with diabetes management,

206