Knowledge Graph-Based Recipe Recommendation for Lifestyle-Related Disease Management

Lifestyle-related diseases can be reduced by making daily dietary choices. Functional components in foods have the potential to provide benefits in this regard. We constructed a knowledge graph that connects the functional components of foods to recipes, and developed a recommendation system to suggest dishes that may help alleviate lifestyle-related diseases. As dietary requirements vary widely depending on specific diseases and individual conditions, these requirements are combined to form a vast probability distribution. Our proposed system uses probabilistic logic programming to generate recommendations that are based on disease-specific dietary requirements, using a knowledge graph of foods and information about the user’s condition. In the proposed method, nodes in the knowledge graph, such as food functionality or recipes, are characterized by their relationships with other surrounding nodes such as ingredients. We conducted an experiment to suggest recipes specifically for patients with diabetes and dyslipidemia. As a result, we were able to develop a system that accurately suggests recipes to prevent and improve each LRD, tailored to the user’s condition.