MIDS Capstone Project Spring 2016

Predicting Demand of Beer Styles

GOAL:

Help brewers make the right beer in the right place at the right time

But Why Beer Models?

  • Big industry
  • Major year-over-year growth
  • Few players in the data/beer space
  • New players need data to succeed and represent an opportunity

Data Science Approach

Get data:

  • Sales data and consumption data from various sources
  • Untappd, OpenBeerDB, FRED, Twitter

Clean and transform data:

  • Load all data into uniform RDBMS

Explore:

  • Do exploratory analysis to understand correlations and trends

Model:

  • Build and test various models to predict ratings, sentiment, and consumption location, including multiple models by style and location

Present and polish:

  • Visualize models in a Shiny dashboard for breweries to use

Next Steps

  • Build out website with interactive dashboard based on demo
  • Add sentiment aggregated by style to supplement maps
  • Adapt style model to account for regional biases in data
  • Keep looking for new data sources to incorporate
  • Get more feedback from breweries of what problems they have that data science can solve

Last updated:

March 30, 2017