brew-data-analysis

The brew-data-analysis project is a toolkit for analyzing brewing-related data. It enables the exploration, modeling, and visualization of data concerning brewing processes, ingredients, and final product characteristics.

Table of Contents

• About The Project
• Analysis Highlights
• Directory Structure
• Prerequisites
• Installation
• Usage
• Features
• Authors

About The Project

The goal of this project is to analyze homebrewing data to uncover trends and build predictive models. The analysis is based on a comprehensive public dataset of beer recipes sourced from Brewer's Friend.

The project utilizes Python and Jupyter Notebooks to process this data, perform exploratory analysis, build a machine learning model to predict beer styles, and generate insightful visual reports.

Analysis Highlights

Our analysis is detailed across two primary reports, uncovering key insights from the dataset:

Report 1: Exploratory Data Analysis (EDA)

This report focuses on cleaning, understanding, and visualizing the core components of the dataset. Key findings include:

• Data Cleaning: Preprocessing steps to handle missing values and standardize data for accurate analysis.
• Parameter Distributions: Visualizations of key beer attributes like Alcohol By Volume (ABV), International Bitterness Units (IBU), and Standard Reference Method (SRM) for color, revealing common ranges and outliers in homebrewing.
• Correlation Matrix: Analysis of the relationships between different brewing parameters (e.g., how Original Gravity relates to ABV).

Report 2: Predictive Modeling & Style Analysis

This report builds on the EDA by applying machine learning to understand and predict beer styles using two distinct approaches:

• Clustering (Unsupervised Analysis): We first used clustering algorithms to group recipes based on their inherent similarities (e.g., SRM, ABV, IBU) without using the pre-defined style labels. This exploratory step helps identify natural groupings within the data and reveals how they align with established categories.
• Classification (Supervised Modeling): A classification model was then trained to predict the official BJCP (Beer Judge Certification Program) style of a beer based on its recipe features. This is the primary predictive task of the project.
• Model Evaluation: The report includes a detailed classification report and a confusion matrix to assess the classification model's performance, identifying which beer styles are most accurately predicted.

Directory Structure

• data/ – Contains the raw and processed Brewer's Friend dataset.
• model/ – Holds analytical scripts and trained machine learning models.
• raports/ – Stores the Jupyter Notebooks (raport_1.ipynb, report_2.ipynb) and generated visualizations.
• utility/ – Includes helper functions for data loading and cleaning.
• main.py – The main script to execute the full analysis pipeline.
• constants.py – Stores constants and configuration settings for the project.

Prerequisites

• Python 3.7+
• A requirements.txt file is provided. Key libraries include:
  • pandas
  • numpy
  • matplotlib
  • seaborn
  • scikit-learn
  • jupyter

Installation

1. Clone the repository:

   git clone https://github.com/varrios/brew-data-analysis.git
   

2. Navigate to the project directory:

   cd brew-data-analysis
   

3. Install the required packages:

   pip install -r requirements.txt
   

Usage

• To run the main analysis script from your terminal:

  python main.py
  

• For interactive analysis, you can run the Jupyter Notebooks located in the reports/ directory:

  jupyter notebook
  

  Then, open raport_1.ipynb or raport_2.ipynb.

Features

• Loading and preprocessing of the Brewer's Friend dataset.
• Exploratory Data Analysis (EDA) and rich visualization of key brewing parameters.
• Building and evaluating predictive models for clustering and classifying beer styles.
• Generating summary reports and visualizations to communicate findings.

Authors

• @Anna Sztukowska
• @Michał Sugalski
• @Lucjan Gackowski