AI Model Defining Heart Healthy Foods

Overview

Project Number: 081-670-81 Title: AI Model Defining Heart Healthy Foods Team Member(s): Wanqi Li Purpose: This project builds an AI-assisted tool that helps people identify heart-healthy foods using objective nutrition data. The goal is to combine evidence-based heart-health guidelines with interpretable machine learning so the decision logic is transparent and easy to validate. Procedure: Nutrition records were collected from the USDA FoodData Central database and standardized per 100 g. Key features included calories, saturated fat, total sugar, fiber, and sodium. A rule-based “ground truth” label was engineered using percent Daily Value for fiber and sodium plus saturated fat as a percent of calories. Two rule groups were implemented to better handle different food types: Group 1 (naturally fiber-rich foods) required SatFat_PctCalories < 6% and Sodium_DV% ≤ 10%; Group 2 (protein foods) required SatFat_PctCalories ≤ 10% and Sodium_DV% ≤ 10%. Data cleaning removed duplicate food names and validated unit consistency. A shallow decision tree model (max depth 4) was trained to learn the labeling logic and provide an interpretable classifier for fast screening. Observations/Data/Results: After cleaning, 7,044 foods remained with zero duplicate names. The decision tree achieved accuracy 0.966, precision 0.925, and recall 0.979 on a held-out test split. Changing from a single-rule system to the two-group system converted 450 foods from No→Yes while preserving safety constraints (0 No→Yes flips in Restaurant/Fast Foods). Conclusions: An interpretable AI pipeline can reliably classify foods by heart-health guidelines, highlight tradeoffs in thresholds across food groups, and support safer everyday food choices.

Competition history

• CSEF 2026 Medicine & Physiology (Junior Division) · Entry J-15-03

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• All Biomedical and Health Sciences projects
• CSEF Biomedical and Health Sciences projects
• Every CSEF 2026 project