Structure of Set of Association Rules Based on Concept Lattice

Abstract

It is important to propose effective algorithms that find basic association rules and generate all consequence association rules from those basic rules. In this paper, we propose the new concept of eliminable itemset to show how to represent itemset by generators and eliminable itemsets. Using algebraic approach based on equivalence relations, we propose a new approach to partition the set of association rules into basic and consequence sets. After describing their strict relations, we propose two ways to derive all consequence association rules from the basic association rules. These two ways satisfy the properties: sufficiency, preserved confidence. Moreover, they do not derive repeated consequence rules. Hence, we save much time for discovering association rule mining.