DFR-HL: Diabetic Food Recommendation Using Hybrid Learning Methods

Abstract

Diabetes affects a large number of people in modern culture. Individuals must keep track of food calories and total calories consumed daily to maintain a balanced diet. Type 2 diabetes is a devastating metabolic illness that may manifest in many symptoms and complications throughout the body. In the modern day, diabetics may be found throughout all age groups in society. The increased number of reported diabetes patients may be attributed to different causes, including but not limited to harmful or chemical components blended into the food, obesity, working culture and improper diet plan, atypical lifestyle, consuming food habits, and environmental variables. As a result, saving human life requires a proper diagnosis of diabetes. When used in the healthcare industry, machine learning techniques may help doctors foresee the onset of diabetes and other complications. This research proposes the Diabetic Food Recommendation System (DFR-HL) to identify diabetes and advice patients on managing their condition via diet (DFRS). The datasets are normalized using a standard scalar with an improved Decision Tree (IDT), and the feature is selected using a Random forest. Finally, the classification has been done with Hybrid (CNN with Resnet50) DL algorithms. The experimental results are compared with performance metrics.