Inductive Databases and Multiple Uses of Frequent Itemsets: The cInQ Approach