Uncertainty in Concept Hierarchies for Generalization in Data Mining

Abstract

Attribute oriented induction is an approach used in data mining to provide summaries of data in a database by the process of generalization that can be used for knowledge discovery in the form of rules or patterns. This is accomplished through the use of a concept hierarchy. When uncertainty is involved in the development and use of the concept hierarchy, the theory behind the uncertainty models in use must first be established. This chapter focuses on providing the foundations for defining imprecise hierarchies and the generalization process with crisp and rough data and hierarchies. Scaling and efficiency issues here involve the problems of creation of appropriate concept hierarchies and the scaling of the generalization process to deal with large databases.