Unsupervised Learning

Abstract

For supervised learning, we have one or more targets we want to predict using a set of explanatory variables. But not all data analysis consists of making prediction models. Sometimes we are just trying to find out what structure is actually in the data we analyze. There can be several reasons for this. Sometimes unknown structures can tell us more about the data. Sometimes we want to explicitly avoid an unknown structure (if we have datasets that are supposed to be similar, we don’t want to discover later that there are systematic differences). Whatever the reason, unsupervised learning concerns finding unknown structures in data.