Unsupervised Learning: Clustering

In contrast to supervised learning, unsupervised learning fits a model to observations assuming there is no dependent random variable, output, or response. That is, a set of input observations is gathered and treated as a set of random variables and analyzed as is. None of the observations is treated differently from the others. An informal way to say this is that there is no Y. For this reason, sometimes classification data that includes the Y as the class is called labeled data but clustering data is called unlabeled. Then, it’s as if the task of clustering is to surmise what variable Y should have been measured (but wasn’t). Another way to think of this is to assume that there are n independent data vectors (X ₁, ...,X_p,Y) but that all the Y _is are missing, and in fact someone has even hidden the definition of Y.