I love beer, especially craft beers, and I also love what people say about beers on Twitter. One evening I had a wild idea to see if I can find out what are the popular types of beers (pilsners, stouts, etc) based in a particular region of the United States. I wanted to know if Pilsners are popular in the West and if Ales are popular in the South, etc.
Using Zapier, I extracted tweets using the #craftbeer search term and saved the them, along with the location of the tweeter (if available), then did some ETL in Rapidminer, harnessed our text mining capabilities, and then built a simple decision tree.
I should give you plenty of warning here because this is an ongoing personal project of mine, and my sample size is ridiculously small. Out of 2,000+ tweets, I ended up with about 80 tweets after I cleaned out the noise and looked for words like stout, ale, etc AND that had a US location available. So take this for what it’s worth, an interesting exercise in mining Twitter, with an emerging interesting result. Maybe.