The goal of this project was to develop and deploy a working recommender to help people choose what whiskey or beer you should drink based on your preferences.

In order to get reviews, to run natural language processing on, I scraped Beer Advocate and the LA Whiskey Society. After cleaning the data, I built a model using Count Vectorizer with a ngram of (2, 2) and then passed the data through Truncated SVD in order to reduce the dimensionality to 5 features. Once I had this reduced feature space, I performed cosine similarity to determine which whiskies and beers are most similar to each other. My recommender then goes through this list and pulls out the top 3 most similar whiskies or beers for your selection.

If you would like to get a recommendation on what whiskey or beer you should drink next, feel free to head to whatshouldidrink.xyz.
--> A whiskey recommendation will be based on the beer you select and your beer recommendation will be based on the whiskey you select.

I ran my scrapers on an AWS EC2 instance.
The beer scraper can be found here and the whiskey scraper can be found here.

My exploratory data analysis on both the whiskey and beer reviews can be found here.

The code for testing and building out my model is here.

The code for testing and building my recommender is here.

All the files for the flask application can be found here.
The python code is here.
The html files are here.

If you would like to check out the PowerPoint presentation click here.