Association rules... and what’s next? — Towards second generation data mining systems

In the process of rule generation from databases, the volume of generated rules often greatly exceeds the size of the underlying database. Typically only a small fraction of that large volume of rules is of any interest to the user. We believe that the main challenge facing database mining is what to do with the rules after having generated them. Rule post-processing involves selecting rules which are relevant or interesting, building applications which use the rules and finally, combining rules together to form a larger and more meaningful statements. In this paper we propose an application programming interface which enables faster development of applications which rely on rules. We also provide a rule query language which allows both selective rule generation as well as retrieval of selected categories of rules from the pre-generated rule collections.