Extracting minimal non-redundant association rules from QCIL

The association rules are usually extracted from frequent itemsets (FIs) but the number is enormous, and there are some FIs with the same tidset, so many redundant rules exist in the mined rules. The frequent closed itemsets are adopted in order to reduce the number of FIs and without lose any useful information. With the inherent closure properties in objects and attributes, the concept lattice is very suitable for representing the relation between closed itemsets. In this paper, the quantitative closed itemset lattice(QCIL) is formed with the modified node structure, and a scheme of building QCIL incrementally from database is adopted. The minimal non-redundant rules are that with minimal antecedent and maximal consequent, and they correspond to the set of the least itemsets (SLIT) of the nodes in QCIL. An innovative algorithm of extracting the association rule by using the SLIT is presented, which can directly extract minimal non-redundant association rules from QCIL.