Data Mining with Concept Generalization Digraphs

Wanlin Pang, Robert J. Hilderman, Howard J. Hamilton, and Scott D. Goodwin Department of Computer Science University of Regina Regina, Saskatchewan, Canada, S4S 0A2 fpang,hilder,hamilton,[email protected] Abstract We present the Path-Based Generalization and Bias-Based Generalization algorithms for attribute-oriented induction using concept generalization digraphs. Concept generalization digraphs are constructed from multiple concept hierarchies associated with an attribute, or set of attributes, in a database. The algorithms e ciently manage the generalization process by avoiding unnecessary re-generalization and by determining which intermediate generalized relations to store for possible future use.