Recipe Recommendation System

Table of Contents

• Overview
• Features
• Technologies Used
• Dataset
• Architecture
• How It Works
  • Recommendation by Ingredients
  • Recommendation by Recipe Name
  • Recommendation by Nutrition
• Acknowledgments

Overview

The Recipe Recommendation System is an intelligent web application that assists users in finding recipes based on their preferences. It utilizes Apache Spark's powerful data processing capabilities to recommend recipes based on various inputs: ingredients, recipe names, or nutritional values. The system is designed for flexibility and scalability, making it suitable for handling large datasets efficiently.

Features

• Search by Ingredients: Input ingredients to get a list of recipes that utilize those ingredients.
• Search by Recipe Name: Enter a recipe name to find similar recipes based on their ingredients.
• Search by Nutrition: Filter recipes based on specific nutritional requirements (fat, protein, carbohydrates).
• Full Recipe Details: Each recommendation includes detailed information such as ingredients, nutritional values, cooking directions, and preparation time.
• User-Friendly Interface: A simple web interface that allows users to easily navigate through options and view recommendations.

Technologies Used

• Python: The main programming language used for backend development.
• Apache Spark: Utilized for big data processing, enabling fast and scalable computations.
• Streamlit: A lightweight web framework that provides the backbone for the web application.
• HTML/CSS: Used for the frontend interface design.
• CSV: The dataset is stored in a CSV format, making it easy to read and manipulate.

Dataset

The dataset used for this project is a CSV file named df_cleaned.csv, which contains various details about recipes. Key columns in the dataset include:

• recipe_name: The name of the recipe.
• ingredients: A list of ingredients required for the recipe.
• nutrition: Nutritional information, including total fat, protein, and carbohydrates (in grams).
• directions: Step-by-step instructions for preparing the recipe.
• timing: The time required to prepare and cook the recipe.

Note: Make sure to place the dataset file in the root directory of the project.

Architecture

The architecture of the application is structured as follows:

1. Data Ingestion: The application reads the recipe data from a CSV file into a Spark DataFrame.
2. Data Processing:
   • Tokenization: Ingredients are tokenized into words for better analysis.
   • TF-IDF Transformation: The Term Frequency-Inverse Document Frequency (TF-IDF) method is applied to quantify the importance of each ingredient in the dataset.
3. Recommendation Algorithms:
   • Centered Cosine Similarity: Utilized for calculating similarities between recipes based on user input.
4. Web Application: The Streamlit web server handles user requests and serves HTML templates to display the results.

Prerequisites

Before running the project, ensure you have the following installed:

• Python 3.x
• Apache Spark (with PySpark installed)
• Streamlit
• Other required libraries (see requirements.txt for the full list)

How It Works

Recommendation by Ingredients

When users enter ingredients, the following steps are executed:

Tokenization: The ingredients are tokenized into individual words for analysis.

TF-IDF Transformation:

The tokenized words are transformed into a TF-IDF representation, which helps determine the importance of each ingredient in the context of the entire dataset.

Centered Cosine Similarity Calculation:

The application calculates the centered cosine similarity between the user's input and each recipe in the dataset.

Display Recommendations:

Recipes with a similarity score above a defined threshold are displayed, along with their full details.

Recommendation by Recipe Name

For searching by recipe name:

The application filters the dataset for recipes that contain the input recipe name. The ingredients of the found recipe are processed similarly to the ingredient-based recommendation. Similar recipes are identified using centered cosine similarity, and recommendations are displayed.

Recommendation by Nutrition

For nutritional recommendations:

The user inputs desired nutritional values for fat, protein, and carbohydrates. The application filters recipes based on these values, allowing a small margin for variability. Recommended recipes are displayed with their full details.

Acknowledgments

Special thanks to the open-source community for the tools and libraries used in this project. Thanks to the dataset providers for making recipe data available.