TheCocatailEDA

This project focuses on exploratory data analysis (EDA) and clustering of cocktail recipes, based on a dataset from TheCocktailDB. See exac changes in changelog

Table of Contents

1. Introduction
2. Installation and following
3. Dataset
4. EDA Conclusions
5. Silhouette Score Results

1. Introduction

This project performs EDA and clustering on a dataset of cocktails and their ingredients. The dataset is available in the data/ folder.

Project Structure

project-root/
├── configs/                             # Configuration files for analysis and preprocessing
│   ├── analysis_configs/                # Analysis-specific configuration files
│   │   ├── general_analysis_config.yaml
│   │   ├── ingredient_analysis_config.yaml
│   │   └── tag_analysis_config.yaml
│   ├── preprocessing_configs/           # Preprocessing-specific configuration files
│   │   ├── data_simplification_config.yaml
│   │   ├── tagging_config.yaml
│   │   └── global_configs.yaml
├── data/                                # Data directory
│   ├── processed/                       # Processed dataset
│   │   ├── processed_cocktail_dataset.json
│   │   └── clustered_cocktail_dataset.json
│   └── raw/                             # Raw dataset (add any initial datasets here)
├── notebooks/                           # Jupyter notebooks for data exploration and visualization
├── outputs/                             # Directory for any output files or results
├── src/                                 # Source code
│   ├── clustering/  
│   │   └── clustering.py
│   ├── analysis/                        # Analysis-related scripts
│   │   ├── general_analysis.py
│   │   ├── ingredients_analysis.py
│   │   └── tag_analysis.py
│   ├── preprocessing_scripts/           # Preprocessing scripts
│   │   ├── one_hot_encode_tags.py
│   │   ├── simplify_data.py
│   │   └── tagging_script.py
├── .gitignore                           # Git ignore file
├── CHANGELOG.md                         # Project changelog
├── environment.yaml                     # Conda environment setup file
├── pyproject.toml                       # Project configuration file
├── README.md                            # Project README with documentation
└── requirements.txt                     # Python dependencies

2. Installation

Create Conda Environment

1. Clone this repository:

   git clone https://github.com/Marcelele-0/TheCocktailEDA
   cd TheCocktailEDA

2. Create and activate the Conda environment:

   conda env create -f environment.yaml
   conda activate cocktail-clustering

3. Install dependencies:

   pip install -r requirements.txt

Usage

• Make sure to set global config to use processed data.

• You can enable and run analysis by choosing interesting functions in configs and run them with:

    python src/analysis/general_analysis.py
    python src/analysis/ingredients_analysis.py
    python src/analysis/tag_analysis.py

• You can preprocess data by running (configs are set up correctly by default - make sure global config is set to processed data if you changed it in analysis):

    python src/preprocessing_scripts/simplify_data.py
    python src/preprocessing_scripts/tagging_script.py
    python src/preprocessing_scripts/one_hot_encode_tags.py

• Now cluster the data by (config is not yet set up so it will inform about everything):

    python src/clustering/clustering.py 

3. Dataset

The dataset is stored in JSON format in the data/ folder.

4. EDA Conclusions

- Data shema

Cocktail

• id (integer): Unique identifier of the cocktail
  Example: 11000, 11001

• name (string): Name of the cocktail
  Example: "Mojito", "Old Fashioned"

• tags (list) - updated in tagging_script.py Example: "tags":["IBA","Classic","Alcoholic","Expensive","Savory"]

• instructions (string, nullable): Instructions on how to prepare the cocktail
  Example: "Muddle mint leaves with sugar and lime juice..."

• alcoholic (boolean): Indicates whether the cocktail contains alcohol
  Example: true, false

• category (enum, nullable): Category of the cocktail
  Example: "Cocktail", "Ordinary Drink", "Shot"

• glass (enum, nullable): Type of glass used for serving
  Example: "Highball glass", "Old-fashioned glass"

• imageUrl (string, nullable): URL of the cocktail image - deleted in simplify_data.py Example: https://cocktails.solvro.pl/images/cocktails/mojito.png

• createdAt (string, nullable): Record creation date - deleted in simplify_data.py Example: "2024-08-19 18:39:58"

• updatedAt (string, nullable): Record update date - deleted in simplify_data.py Example: "2024-08-21 10:12:58"

• one_hot_tags (list) - created in one_hot_encode_tags.py Example: "oneHotTags": [1, 0, 1, ..., 0, 1] # Represents presence of tags in a binary format

Ingredients

• id (integer): Unique identifier of the ingredient
  Example: 10, 11

• name (string): Name of the ingredient
  Example: "Red wine", "Grapefruit juice"

• description (string, nullable): Description of the ingredient
  Example: "Red wine is a type of wine made from dark-colored grape varieties..."

• alcohol (boolean, nullable): Indicates whether the ingredient contains alcohol
  Example: true, false

• type (enum, nullable): Type of ingredient
  Example: "Vodka", "Gin", "Juice"

• percentage (number, nullable): Alcohol percentage of the ingredient
  Example: 40, null

• imageUrl (string, nullable): URL of the ingredient image
  Example: https://cocktails.solvro.pl/images/ingredients/rose.png

• createdAt (string, nullable): Record creation date for the ingredient
  Example: "2024-08-19 18:39:58"

• updatedAt (string, nullable): Record update date for the ingredient
  Example: "2024-08-21 10:12:58"

• measure (string): Measurement or quantity of the ingredient used
  Example: "1/2 oz"

• Unique ingredients in the dataset:

  ['Soda water', 'Light Rum', 'Lime', 'Mint', 'Sugar', 'Water', 'Angostura Bitters', 'Bourbon', 'lemon', 'Vodka', 'Gin', 'Tequila', 'Coca-Cola', 'Sweet Vermouth', 'Campari', 'Powdered Sugar', 'Blended Whiskey', 'Cherry', 'Dry Vermouth', 'Olive', 'Lime Juice', 'Salt', 'Triple Sec', 'Ice', 'Maraschino Cherry', 'Orange Peel', 'Ginger Ale', 'Apricot Brandy', 'Lemon Juice', 'Amaretto', 'Sloe Gin', 'Southern Comfort', 'Lemon Peel', 'Orange Bitters', 'Yellow Chartreuse', 'Creme De Cacao', 'Light Cream', 'Nutmeg', 'Brandy', 'Lemon vodka', 'Pineapple Juice', 'Blackberry Brandy', 'Kummel', 'Dark Rum', 'Egg White', 'Kahlua', 'Club Soda', 'White Creme de Menthe', 'Tea', 'Whipped Cream', 'Apple Brandy', 'Applejack', 'Orange', 'Benedictine', 'Wine', 'Champagne', 'Green Creme de Menthe', 'Grand Marnier', 'Bitters', 'Scotch', 'Banana', 'Carbonated Water', 'Coffee Liqueur', 'Celery Salt', 'Tabasco Sauce', 'Tomato Juice', 'Worcestershire Sauce', 'Blue Curacao', 'Lemonade', 'Anejo Rum', 'Orange Juice', 'Tia Maria', 'Maraschino Liqueur', 'Grenadine', 'Egg', 'Cachaca', 'Egg Yolk', 'Cognac', 'Cherry Brandy', 'Port', 'Chocolate Ice-cream', 'Dubonnet Rouge', 'Sugar Syrup', 'Pineapple', 'Tonic Water', 'Orange spiral', 'Strawberries', 'Heavy cream', 'Galliano', 'Irish Whiskey', 'Peach brandy', 'Sweet and Sour', 'Green Chartreuse', 'Drambuie', 'Orgeat Syrup', 'Grapefruit Juice', 'Red Wine', 'Raspberry syrup', 'Sherry', 'Coffee Brandy', 'Lime vodka', 'Lemon-lime soda']

• Unique tags in the dataset:

  ['IBA' 'ContemporaryClassic' 'Alcoholic' 'USA' 'Asia' 'Vegan' 'Citrus' 'Brunch' 'Hangover' 'Mild' 'Classic' 'Expensive' 'Savory' 'Strong' 'StrongFlavor' 'Vegetarian' 'Sour' 'Christmas' 'Beach' 'DinnerParty' 'Summer' 'Chilli' 'Dairy' 'Nutty' 'Cold' 'Fruity' 'Breakfast' 'NewEra']

  Approach to Tagging and Ingredients

  In our cocktail analysis project, we have adopted a structured and dynamic approach to tagging cocktails based on their ingredients. This system not only enhances the organization of our data but also facilitates deeper insights into the relationships between different cocktails and their components. The coctails will be clustered based on tags.

  Tagging System

  Our tagging framework utilizes a set of predefined tags that categorize cocktails based on their ingredient composition. Each tag has specific criteria that must be met, allowing for a flexible and dynamic assignment of tags. The key components of our tagging system include:

  • Tag Definitions: Tags are defined in a configuration file using YAML format. Each tag has associated ingredients and a threshold that determines how many of those ingredients must be present in a cocktail for the tag to be assigned. This approach allows for easy modifications and additions to the tagging rules as our understanding of cocktails evolves. Tags are defined in a YAML configuration file (tagging_config.yaml).
  • Ingredient Categorization: Ingredients are categorized into various groups, such as strong, new era, classic, and regional ingredients. This classification helps in understanding the characteristics of cocktails and their flavor profiles.
  • Dynamic Assignment: The tagging mechanism dynamically assigns tags based on the ingredients present in each cocktail. This means that as we expand our ingredient database or modify our tagging criteria, the tagging process remains adaptable and robust.

Silhouette Score Results

At the beginning of the project, the Silhouette Score was approximately 0.18, indicating relatively low clustering quality. Consequently, several iterations of modifications were made to the tags and data to enhance the results.

Actions Taken

1. Disabling Certain Tags: It was determined that disabling the assignment of specific tags allowed for clearer grouping. The following main tags were turned off:

   • Classic
   • Contemporary Classic
   • New Era

2. Deleting Tags Completely: Some tags were found to be overly homogeneous or too specific, leading to their removal. For example 'Chili' - low presence 'Alcoholic' - too high presence

3. Using MinMaxScaler to appply weights.

Result: The Silhouette Score increased to above 0.3, indicating a significant improvement in clustering quality and better differentiation between groups.

These changes were implemented to achieve more coherent and interpretable clusters while enhancing the readability of the results and the accuracy of the cocktail grouping.