Extended MRI-Cube Algorithm for Mining Multi-Relational Patterns

Association rule mining is one of the most important and basic technique in data mining, which has been studied extensively and has a wide range of applications. Two stream of previous work has dealt with the discovery of association rules over multiple relations: prolog databases and datalog queries. The MRI Iceberg-cubes mining method introduces a new perspective. However, it does not take the cyclic join paths into account, therefore, in this paper, we will introduce an algorithm, Extended-MRI-cube, which is based on the MRI-Cube algorithm, to handle the cyclic join path situation. Experiments show it is more applicable and effective than the previous one.