###Every proper club needs a proper club cocktail.
Can we algorithmically find our club cocktail using a large database of cocktail drinks? Yes.
Here are our goals to accomplish:
- Determine the ingredients (and perhaps the directions to make the cocktail, too)
- This may require an objective function.
- One of the data sets (
6k*.csv) has ratings that we can use. - Using the dataset to weed out bad combinations is good too!
- Should have some flair to it.
- More data is better. We should look at popularity too.
- Determine a name for the cocktail (this should be data determined too)
- Naively, gather all names of cocktails, split on whitespace, and randomly pick.
- Better: considering the ingredients we have chosen, look for a pair or triplet of existing cocktails that can be "summed up" to produce our cocktail. This is like a regression task.
- using Google Trends to find "hot names", i.e. "One Miley Direction" would be hilarious.
- Nearest Neighbour approaches?
- Look for interesting relationships in the dataset. "orange juice => vodka"? Some tools that might help: association rules, network analysis
- Create a cool postcard-sized graphic about our cocktail (with recipe). We'll use this to create actual postcards to promote our club: "Dear P.M. Harper, I've noticed you could use a drink...". "Dear Ottawa Citizen, we have provided for you a result you might finding interesting-- the url is at ...."
- Can we get a local drinking establishment to serve it? Maybe exclusively to our members/meetings?
- Can we light it on fire? Why not?
- Actually drink it. Our next meeting will be supplied with our new Data Origami cocktail.