K-Means Clustering: Unsupervised Learning Applied on Magic:The Gathering

Like many other unsupervised learning algorithms, K-means clustering can work wonders if used as a way to generate inputs for a supervised Machine Learning algorithm (for instance, a classifier). There are many ways we could have approached the recommendation problem: given a card, suggest other cards that go well with it, without using any data about the cards except which decks they appear in (that is, no cheating and asking for more data about the cards like color, price, or an expert’s opinion). As you can see, I omit how many times a card appears on a given deck for this part, and just look at the relative number of apparitions for a card on a given cluster.

Source: www.datastuff.tech