NOTE: please use nbviewer to view jupyter notebook files if it does not load. Simply copy the link to the jupyter notebook file, go to this website: https://nbviewer.jupyter.org/ and paste into the box.
The goal of this project was to create a recommender system that will take a beer name as an input and return the top N recommendations based on item similiarity. Content based recommendations was explored in a different notebook but the goal was to return recommendations that are varied and not just more of the same (which is what the content-base recommender would do). For example, if I like IPA beers I want to also explore non-IPA beers that may have similar charteristics but are different enough to be categorized as a different class of beer.
The Data:
- 56 subcategories of beer
- 4964 unique beers
- ~88K unique users
- ~1.4M user-review pairings
Data was scraped from a popular beer review website: https://www.beeradvocate.com/
Main Tools:
Surprise Library - http://surpriselib.com/
scikit-learn: https://scikit-learn.org/stable/