Unsupervised Learning

Abstract

Supervised learning discovers patterns in the data that relate data attributes to a class attribute. These patterns are then utilized to predict the values of the class attribute of future data instances. These classes indicate some real-world predictive or classification tasks such as determining whether a news article belongs to the category of sports or politics, or whether a patient has a particular disease. However, in some other applications, the data have no class attributes. The user wants to explore the data to find some intrinsic structures in them. Clustering is one technology for finding such structures. It organizes data instances into similarity groups, called clusters such that the data instances in the same cluster are similar to each other and data instances in different clusters are very different from each other. Clustering is often called unsupervised learning, because unlike supervised learning, class values denoting an a priori partition or grouping of the data are not given. Note that according to this definition, we can also say that association rule mining is an unsupervised learning task. However, due to historical reasons, clustering is closely associated and even synonymous with unsupervised learning while association rule mining is not. We follow this convention, and describe some main clustering techniques in this chapter.