Incremental refinement of mining queries

Abstract

A first attempt to extract association rules from a database frequently yields a significant number of rules, which may be rather difficult for the user to browse in searching interesting information. However, powerful languages allow the user to specify complex mining queries to reduce the amount of extracted information. Hence, a suitable rule set may be obtained by means of a progressive refinement of the initial query. To assist the user in the refinement process, we identify several types of containment relationships between mining queries that may lead the process. Since the repeated extraction of a large rule set is computationally expensive, we propose an algorithm to perform an incremental recomputation of the output rule set. This algorithm is based on the detection of containment relationships between mining queries.