ML beer recommendation engine

[scott-eddy]

So i think this would be best done by @brigidangelini, but I think it would be awesome to train up some ML model and have a submission in an upcoming FC be from a draper gang hive mind. I don't know how to do this, but it would be cool.

Also, in a general sense if other people were to ever use this I think they would want a "club recommendation" as well.

[drc0308]

I too have thought about this and even had some crappy one. It's a hard problem but now we have some cleaned data at least haha.

[brigidangelini]

I can see this being a couple of different things off the top of my head

• an engine that lets you input the beer you're suggesting bringing, the output could be something along the lines of how many votes the model thinks the beer might score, or a proxy for this which i think is what @drc56 flight club bitterness units was getting at back in the day, something alone those lines. this would be doable with all the data we have available.

• a true recommendation engine that just serves you some random beer as the output and how it would do at a flight club. maybe this pings beer advocate to pick the random beer and outputs what it guesses the number of votes would be. this doesn't seem that useful though because there are tons of beers on BA that are no longer available/will never be available in NE/etc

• flight club similarity scorer: similar to the first option, it tells you what beer we've had at flight club is most similar to the beer you input to the engine, maybe a ranked list of the most similar beers

we can probably do some pretty fun stuff. I really like the idea of having something that would be able to pick a beer that someone could buy and serve it at a future FC. I'll think about smart ways to pull in outside data sources for this sort of thing (or if anyone else has ideas about this, it doesn't have to be limited to BA's api, we could use conventional web scraping methods etc)