Skip to content

Modified version of Le Wagon Data Science capstone project (March 2024)

Notifications You must be signed in to change notification settings

tjyana/tjyana-mealmuse

Repository files navigation

Access the app here: https://tjyana-mealmuse.streamlit.app/

This app was created in tandem with 3 teammates from my cohort at Le Wagon.

MealMuse: From what you have to what you crave

・About MealMuse: MealMuse is an innovative AI-driven recipe generator designed to inspire your culinary creativity. By simply inputting your available ingredients, MealMuse crafts unique, palatable recipes complemented by visually appealing images, ensuring an exciting cooking adventure every time.

・How It Works:

  1. MealMuse utilizes an advanced pairing algorithm to determine compatible ingredients, forming the basis of potential recipes. These recipes are then evaluated through a sophisticated AI model to ensure they meet our quality standards. This iterative process, supported by a vast dataset from food.com, guarantees the delivery of enjoyable and innovative recipes.
  2. Utilizes over 170,000 ingredients and 180,000 recipes from food.com for comprehensive pairing options.
  3. Employs a robust AI model with a 67% accuracy rate in predicting recipe success. Continuously refine recipes to meet high-quality standards.

・Dataset and Model Information Dataset: Over 170,000 ingredients, 180,000 recipes, and 1.1 million user ratings from food.com. Model Building: Feature selection, recipe rating encoding, and threshold fine-tuning. Evaluation with multiple scikit-learn models and enhancement through Google BERT tokenizers.