Recommander-System-

This project implements a deep learning-based two-tower recommendation system. The model takes user nutritional requirements as input and recommends the most suitable food items. The system is built with PyTorch and trained using paired user-food data.

Obtaining Dataset

Unzip the datset in same directory to get a combined_Dataset.csv

• Do not push the csv dataset into the remote repository (size limit exceeded)
• For local reference only

Project Structure

├── recommander.py           # Model training script
├── interact.py              # Interactive recommendation script
├── FOOD-DATA-GROUP-Prossced.csv  # Original food dataset
├── combined_Dataset.csv      # User-food paired dataset (for training)
├── trained_model.pth         # Trained model
├── scaler_user.pkl           # Standard scaler for user features
├── scaler_food.pkl           # Standard scaler for food features
├── README.md                 # Project description

Features

1. Model Training:

• recommander.py defines the two-tower model (TwoTowerModel), which generates embeddings for user and food features.
• Uses combined_Dataset.csv to train the model.
• Saves the trained model and scalers upon completion.

2. Recommendation Functionality:

• Interact.py provides an interactive command-line tool where users can input their nutritional requirements to receive food recommendations.
• Recommendations are generated using FOOD-DATA-GROUP-Prossced.csv.

3. Data Processing:

• User and food features are standardized using StandardScaler.
• Combined_Dataset.csv contains paired user-food data with labels (indicating whether the food meets the user's requirements).

Requirements

Install the following dependencies:

• Python 3.8+
• PyTorch
• scikit-learn
• pandas
• numpy
• joblib

Datasets

1. Original Food Dataset (FOOD-DATA-GROUP-Prossced.csv) This dataset contains nutritional features for each food item, including:

   • Caloric Value
   • Fat
   • Carbohydrates
   • Sugars
   • Protein
   • Dietary Fiber
   • Cholesterol
   • Sodium

2. Paired Dataset (combined_Dataset.csv) This dataset pairs user features with food features and includes a label (Label) indicating whether the food meets the user's nutritional requirements:

   • User features (e.g., caloric need, minimum protein requirement, etc.)
   • Food features (e.g., calories, fat, sugar, etc.)
   • Label (1 for a match, 0 for a mismatch)

How to Use

1. Train the Model Run recommander.py to train the model and save the trained parameters:

   ├── python recommander.py
   

   After training, the following files will be generated:

   trained_model.pth: Saved model parameters.

   scaler_user.pkl: User feature scaler.

   scaler_food.pkl: Food feature scaler.

2. Use the Recommendation System Run interact.py for an interactive recommendation system:

   ├── python interact.py
   

   Provide your nutritional requirements in the following format (comma-separated):

   Enter the following format:
   User ID, User Caloric Value, User Fat Limit, User Carbohydrates Limit, User Sugar Limit, 
   User Protein Min, User Dietary Fiber Min, User Cholesterol Limit, User Sodium Limit
   
   Example input:
   1, 2000, 70, 300, 50, 60, 30, 200, 1500
   

If you don't want to train the model yourself, you can download the trained_model.pth, scaler_user.pkl, scaler_food.pkl into your directory and run python interact.py.