🥗 Healthy Recipe Analysis
Tech Stack: Python, Pandas, NumPy, Scikit learn, Statsmodels, Seaborn, Plotly
Food labels can guide or mislead. This project investigates whether recipes tagged as “healthy” truly meet common nutritional standards, which is a crucial concern for individuals managing health conditions or dietary restrictions.
Over 80,000 recipes from Food.com were analyzed to compare nutritional attributes like sugar and saturated fat content per calorie between “healthy” tagged and untagged dishes.
The analysis includes hypothesis testing, custom feature engineering such as sugar to calorie ratio, and interactive visualizations using Plotly and Seaborn.
A Random Forest classifier was trained to predict whether a recipe deserves the “healthy” label based on its nutritional profile.
After tuning with GridSearchCV, the model achieved 95 percent precision on test data. A fairness evaluation was also conducted to assess model behavior across diverse nutritional cases.
The project highlights the disconnect between perceived and actual healthiness, and shows how data science can help bridge the gap.