Clustering Scotch Whiskies using Non-Negative Matrix Factorization

I happened upon this article while sipping one of my all time favorite single-malt Scotchs: The Laphroaig 10 Year, a wonderfully peaty selection from Islay. It describes how Singular Value Decomposition can be used as a form of matrix factorization to help cluster Scotch-Whiskies based on several “flavor”-metrics. Each of the whiskies being examined is analyzed based on each of the designated “flavor”-metrics and is assigned a value from 0-4 for each. Using SVD to decompose the matrix in to factors groups these whiskies into clusters of similar flavor-patterns.