A new multi-relational mining method exploiting ER models

It presents an approach based on ER models to solve statistical skew and low efficiency problems in multi-relational data mining. It makes use of the constraints between entity sets and relationship sets of an ER model, and applies extended SQL statistical primitives to produce multi-relational frequent patterns on the interesting attributes, avoiding physical joining among all relational tables. It proposes a new algorithm for mining multi-relational frequent itemsets, and two theorems ensure the correctness of the method to relational databases. Other kinds of work are introduced and analyzed, which are based on different conditions or requirements.