Can AI help you eat healthier?

The new AI system will offer a separate weekly menu rooted in balancing Mediterranean cuisine, taste, nutrition and cultural needs, unless you have nut or dairy allergies.

Research: AI-based nutritional advisory system: technical verification with insights from Mediterranean cuisine. Image credits: Dragon Image/Shutterstock

Chronic or lifestyle diseases such as stroke, heart attack, obesity and type 2 diabetes are on the rise, primarily due to switch diet and physical activity habits. Artificial intelligence (AI) is pushed into services to develop personalized, balanced diet plans for healthy people that meet their tastes and needs.

Recent papers published in the journal Nutrition frontier Presents an AI-based nutritional advisory (AINR) system that creates weekly meal plans from Mediterranean menus, meeting calorie and macronutrient recommendations while enforcing food diversity and dietary rules.

introduction

Most people know that they need to eat a physically active and balanced diet. Work habits, family schedules and financial resources often hinder the adoption of such habits. To separate them from junk food, experts have investigated the feasibility and effectiveness of nutritional advisory systems. They are designed to allow users to choose the best diet to improve long-term compliance and health.

AI is currently being used to create the best meal plan based on the user's diet choices and needs, including allergies, seasonality, cultural factors, and calorie needs. Leading language models (LLMs) like CHATGPT represent new approaches to nutrition AI, but current AINRs use rules-based methods instead.

Traditional AI-based systems

Such nutritional recommendation systems may be categorized as traditional and AI-based systems.

Traditional systems

The first type of system uses four different approaches to plan your meals.

• Combination optimization techniques
• Content-based filtering
• Joint filtering
• Hybrid approach

Combination analysis aims to ensure proper dietary diversity, including all food groups and achieve cost-effectiveness while respecting user preferences. Despite mathematical health, the algorithms used here may not accurately reflect the user's goals and choices. It also narrows the area of ​​food diversity, while oversimplifying nutritional rules. These limitations limit their applicability to actual dietary planning in real life.

Three other types of approaches correlate user preferences and food attributes to food attributes to personalize food suggestions. They use past interactions or user profiles to suggest foods represented in the user profile they are using, or similar foods to others with similar profiles. The hybrid approach aims to blend the best parts of these methods to further optimize and increase the relevance of meal planning.

AI-based systems

The AI-based nutrition recommendation system was designed to overcome the obstacles of too little data in many critical areas, scalability issues, and many critical areas that need to be addressed with cold start. Dietary restrictions (such as allergies and cultural taboos) and health goals, and nutritional content of these weight users. Both knowledge-based and machine learning (ML)-based systems are currently used in such systems.

Knowledge-based systems are represented by a multi-objective optimization (MAOO) approach and diet plan generator (MPG) mechanism, or Protein-Ai-Advisor, which edits weekly meal plans based on a balanced diet of nutrients to suit the needs and dietary rules of the user. Conversely, ML-based platforms have expressed preferences to learn from context and user behavior and come up with personalized meal plans for daily use.

AI-based nutritional recommendations (AINR). AINR supplies user profiles and preferences using principles of healthy eating and foods from the Mediterranean diet and culinary database. The AINR core processes this information to suggest healthy, personalized weekly meal plans to users using expertly validated rules and filtering mechanisms.

About the research

The current study presents an AINR designed to provide weekly meal plans based on Mediterranean diets in either Spanish or Turkish. The system was tested with 4,000 generated user profiles including allergies (dairy products, eggs, nuts, fish) and preferences (halal).

How it works

The aim was to prepare a weekly nutrition plan using a rigorous four-step process.

1. Filter your meals with seasonal and local cuisine
2. Exclude diets that are inconsistent with allergies/likes
3. Generate and acquire daily diet plans for energy/macronutrient targets
4. Build weekly plans and implement food diversity rules (e.g. serving up to 3 fish/week, repeat dishes 3x/week>week)

Importantly, human monitoring is required as autogenerated plans are not verified by a nutritionist.

Research Results

Results showed 100% filtering accuracy of country, allergies, preferences and seasonality. However, weekly plans were generated with only about 90% of the profile.

• Spanish users with milk/nut allergies received a 0% viable plan due to a database gap (for example, there is no Spanish breakfast options for milk allergies users)
• Turkish male profiles showed low accuracy (87% calories, 66% protein)
• Spanish users and Turkish women achieved 98% calorie accuracy and more than 90% macronutrient accuracy

Milk and nut allergies have hindered weekly planning, especially for Spanish users, due to limited database options. AINR systems require dairy replacements and, most importantly, database extensions with actual verification.

Conclusion

The AINR system “promises to promote more balanced nutritional habits”, but requires improvements to the database of allergies and male Turkish users. Researchers will test it with real users with the Switchto Healthy intervention and expand it for family meal plans.