A method for discovering association rules in a database that employs item constraints for extracting desired data relationships from a data base, thereby reducing the execution time of the rule discovery process and increasing the quality of the information returned. Such constraints allow users to specify the subset of rules in which the users are interested. Given a set of transactions D and constraints represented by a boolean expression .beta., the invention integrates the constraints into a selected rule discovery method rather than implementing the constraints as a post-processing step. The invention quickly discovers association rules that satisfy .beta. and have support and confidence levels greater than or equal to user-specified minimum support and minimum confidence levels, and may be implemented even when a taxonomy is present.