Nutrition-Related Knowledge Graph Neural Network for Food Recommendation

Area of Science:

• Computer Science
• Nutrition Science
• Artificial Intelligence

Background:

• Current food recommendation systems often neglect nutritional content, leading to unhealthy and repetitive suggestions.
• This lack of nutritional consideration can negatively impact users' overall health and dietary diversity.

Purpose of the Study:

• To develop a novel nutrition-related knowledge graph (NRKG) method for enhanced food recommendations.
• To address the limitations of existing systems by incorporating nutritional information and promoting healthy eating habits.

Main Methods:

• Developed a nutrition-related knowledge graph (NRKG) method utilizing graph convolutional networks (GCNs).
• Integrated two key components: user nutrition-related food preferences and recipe nutrition components.
• Constructed a heterogeneous graph connecting recipes with similar nutritional profiles and learned from this structure.

Main Results:

• The NRKG method demonstrated superior performance compared to six baseline methods across five metrics on real-world food datasets.
• Achieved performance improvements of 2.8% to 9.7% over the best baseline method.
• Effectively integrated user preferences with nutritional data for more accurate and personalized recommendations.

Conclusions:

• The proposed NRKG method significantly outperforms existing food recommendation systems.
• This approach effectively promotes healthier and more diverse eating habits by considering nutritional content.
• Offers a comprehensive strategy by integrating recipe nutrition with user preferences, unlike hierarchical information propagation methods.