Semantics and syntactic patterns in data

This paper examines the semantics and syntactic views of classical association rule mining. A relational table is considered as a (knowledge) representation of a universe (= the set of real world entities). A pattern is said to be realizable, if there is a real world phenomenon corresponding to it. The central two issues are: Why do unrealizable data patterns appear? How could they be pruned away? For this purpose, the semantics of the original schema are considered. In additions, semantic is included into the knowledge representation of the universe. Based on model theory, two new relational structures, functions and binary relations, are added to represent some additional semantics of the given universe. Association rule mining based on such additional semantics are considered.