NutriGen: Personalized Meal Plan Generator Leveraging Large Language Models to Enhance Dietary and Nutritional...

Area of Science:

• Nutrition Science
• Artificial Intelligence
• Computational Health

Background:

• Balanced diets are crucial for health, but meal planning is challenging due to complexity and time constraints.
• Existing dietary recommendation systems lack adaptability and real-world applicability.
• Personalized recommendations can improve adherence but often require extensive user input.

Purpose of the Study:

• To introduce NutriGen, a large language model (LLM)-based framework for personalized meal plan generation.
• To address limitations of current systems by incorporating user preferences, constraints, and real-world factors.
• To evaluate the accuracy and practicality of LLM-driven personalized nutrition.

Main Methods:

• Developed NutriGen, a framework leveraging LLMs and prompt engineering.
• Integrated a personalized nutrition database with reliable sources like the USDA database.
• Evaluated LLM performance (Llama 3.1 8B, GPT-3.5 Turbo, DeepSeek V3) for accuracy and processing time.

Main Results:

• LLMs demonstrate significant potential in generating accurate, user-friendly food recommendations.
• Llama 3.1 8B and GPT-3.5 Turbo achieved low percentage errors (1.55% and 3.68%) in meeting caloric targets.
• DeepSeek V3 showed lower accuracy (MAE 10.45%) and slower processing, indicating a need for optimization.

Conclusions:

• LLM-based frameworks like NutriGen offer a practical and scalable solution for personalized meal planning.
• NutriGen effectively addresses limitations of existing systems, providing structured and adaptable dietary guidance.
• Further optimization of LLMs is needed for enhanced performance in personalized nutrition applications.