Diet Recommendation System
L. Suresh,
Kanturi Vikas,
Jannapureddy Kalyani,
Guduru Shruthilaya Kiran,
Bitla Uday Kiran
Irregular eating patterns and poor nutritional awareness have led to increased health complications in
modern lifestyles. This paper presents a Diet Recommendation System that generates personalized food
suggestions by directly matching user-defined nutritional requirements with a large-scale recipe dataset.
Unlike traditional systems that depend on user history, the proposed model uses a content-based filtering
approach, eliminating the cold-start limitation. The system applies the K-Nearest Neighbors (KNN)
algorithm with cosine similarity to identify recipes that closely align with user-defined calorie and
macronutrient targets. The application is built using FastAPI for backend processing and Streamlit for
the frontend interface, with Docker ensuring consistent deployment. Experimental results demonstrate
that the system provides recommendations within a ±10% accuracy range while maintaining low
response latency. The approach ensures transparency, scalability, and immediate usability for new users.
