Knowledge Discovery from Unsupervised Data in Support of Decision Making

Publisher Summary Knowledge discovery and data mining (KDD)—the rapidly growing interdisciplinary field that merges database management, statistics, machine learning, and related areas—aims to extract useful knowledge from a large collection of data. Current technology makes it easy to generate large collections of data of different types in different forms. To some extent, in this variety of data sources, KDD methods can be distinguished by the degree of supervision in data. In general, the “degree of supervision” concerns either the supervision of humans or the supervision in data. Although a domain expert plays an important role in the discovery process, discovery methods themselves are independent of the domain expert but dependent on the supervision in data. Most work in knowledge discovery has focused on supervised discovery tasks that aimed to find useful descriptions (patterns or models) of classes of classified instances. Supervised discovery methods are always driven by feedback about the appropriateness of discovered knowledge. In real-world situations, unsupervised discovery task can also arise quite often. This chapter concerns mainly with knowledge discovery in unsupervised data.