menupal 0.1.3

Menu Recommendation Tool

mc-menu-picker

A content-based recommendation system built on McDonald's menu nutrition data. It computes multi-dimensional similarity across numeric, ingredient, and category features and provides the top-N most similar menu items.

✨ Features

Feature	Description
🔢 Numeric Similarity	Cosine similarity calculation based on calories, fat, carbohydrates, and other numeric fields
🥗 Ingredient Similarity	Jaccard similarity calculation based on one-hot encoded ingredient features
📂 Category Similarity	Jaccard similarity calculation based on one-hot encoded category features
⚖️ Weighted Fusion	Combines three similarity matrices with customizable weights (default [0.4, 0.35, 0.35])
💻 Local Recommendation	Run python main.py to randomly select items and list top-K similar items
🚀 RESTful API	Launch FastAPI service via ./scripts/api.sh for external calls

📦 Installation

git clone https://github.com/911218sky/mc-menu-picker.git
cd mc-menu-picker
pip install -r requirements.txt

📊 Data Preparation

Alternative: You can also download the dataset directly - download here.

Step 1: Extract the provided dataset

# Example extraction (adjust path as needed)
unrar x data.rar ./

The program loads raw data and feature matrices, builds or loads the combined similarity matrix, and displays the top-10 recommendations for a random test ID.

🌐 API Service

Start the FastAPI service using the following command:

bash ./scripts/api.sh

Environment Variables

Environment Variable	Description	Default Value
HOST	Bind address	0.0.0.0
PORT	Listening port	3010
PRODUCTION	If set to true, disables /docs and /redoc endpoints	false

📖 API Documentation

After startup, visit http://localhost:3010/docs to explore the Swagger UI interface.

---

🔧 Development Notes

• Similarity Calculation: The system combines three different similarity algorithms for numeric, ingredient, and category features
• Flexible Weighting: Weight ratios for different features can be adjusted according to requirements
• High Performance: Pre-computed similarity matrices provide fast recommendation responses

📝 Usage Examples

Local Recommendation:

python ./src/main.py
# Output: Random item and its top-10 similar recommendations