A new classification system of beer categories and styles based on large-scale data mining and self-organizing maps of beer recipes

TL;DR

This paper presents a data-driven, scalable classification system for beer styles using large-scale recipe data and self-organizing maps, revealing distinct ingredient and fermentation patterns across beer categories.

Contribution

Introduces a novel, objective taxonomy of beer styles based on extensive data mining and machine learning, surpassing traditional sensory-based classifications.

Findings

Identified four major beer superclusters with unique ingredient patterns

Revealed regional and fermentation-based heterogeneity in beer styles

Provided a scalable framework for beer recipe analysis and development

Abstract

A data-driven quantitative approach was used to develop a novel classification system for beer categories and styles. Sixty-two thousand one hundred twenty-one beer recipes were mined and analyzed, considering ingredient profiles, fermentation parameters, and recipe vital statistics. Statistical analyses combined with self-organizing maps (SOMs) identified four major superclusters that showed distinctive malt and hop usage patterns, style characteristics, and historical brewing traditions. Cold fermented styles showed a conservative grain and hop composition, whereas hot fermented beers exhibited high heterogeneity, reflecting regional preferences and innovation. This new taxonomy offers a reproducible and objective framework beyond traditional sensory-based classifications, providing brewers, researchers, and educators with a scalable tool for recipe analysis and beer development. The findings in this work provide an understanding of beer diversity and open avenues for linking ingredient usage with fermentation profiles and flavor outcomes.