OLAM Cube Selection in On-Line Multidimensional Association Rules Mining System

摘要

Mining association rules from large database is a computation intensive task. To reduce the complexity of association discovery, Lin et al. proposed the concept of OLAM (On-Line Association Mining) cube, an extension of Ice-berg cube used to store frequent multidimensional itemsets. They also presented a framework of on-line multidimensional association rule mining system, called OMARS, which relies heavily on the OLAM cubes to provide an OLAP-like association mining environment. This paper is a companion toward the implementation of OMARS. Particularly, we investigate the problem of selecting appropriate OLAM cubes to materialize and store in OMARS. Several properties of the OLAM cube that are useful to the cube selection are presented. We also discuss how to adopt the greedy method to solve the problem under the storage constraint.