Granular computing approach to finding association rules in relational database

Granular computing is a new information‐processing method. The main objective of this paper is to present a granular computing approach to finding association rules in relational databases. Elementary granules are generated by scanning relational database, and granule table structure for storing information granules is established. By keeping attributes in order and referring to granular computing, frequent k‐item sets are gradually generated from frequent 1‐item sets, frequent 2‐item sets, and so on. Corresponding algorithms are proposed and illustrated with a real example. Experiments on real data show that the algorithms can reduce the number of candidate item sets and save the computing time. © 2009 Wiley Periodicals, Inc.