mrpotatocode/COFFEE_COFFEE_COFFEE

end-to-end data science project by yours truly

Coffee & Machine Learning: An Exploration into the Attributes that Define Single-Origin Coffees

Paper Abstract

In this work, a boosted tree model explores whether combinations of unique coffee attributes can predict the tastes of specialty coffees. The attributes of 550 coffees, collected over a period of four months, include the coffee's country of origin, variety (subspecies), processing method, harvest characteristics (like altitude and season of harvest), and tastes generated through consistent cupping practices. These defining attributes are used as predictors for 22 distinct tasting groups and then compared to actual specialty coffees and found to often match two out of every three tasting groups provided by coffee roasters.

Current Roadmap

• Drink a lot of coffee
• Plan the project
• Put the project on GitHub
• Learn webscraping with rvest
• Collect data (plural)
• Learn GitHub Actions
• Deploy GitHub Actions for automated and scheduled webscraping
• Learn Shiny
• Deploy my first Shiny App
• Learn tidymodels
• Write my first Datasheet
• Write my first Modelcard
• Deploy models
• Deploy Predictive Shiny App
• Write paper
• Build Github pages for the project
• Have an interesting README about the project
• Drink a lot of coffee