An Evolutionary Method for Exceptional Association Rule Set Discovery from Incomplete Database

摘要

A method for exceptional association rule set mining from incomplete database is proposed to discover interesting combination of items in incomplete database. The rule set is defined as each itemset X, Y has weak or no statistical relation to class C, respectively, however, the join of X and Y has strong relation to C. The method extracts the rule set directly as the combination of three rules even though the database has missing values. The method has been developed using a basic structure of an evolutionary graph-based optimization technique and adopting a new evolutionary strategy to accumulate rule sets through its evolutionary process. The method can realize the association analysis between two classes of the incomplete database using chi-square values. We evaluated the performance of the proposed method for exceptional association rule set mining from the incomplete database. The results showed that the method has a potential to realize association analysis in medical field based on the rule set discovery. In addition, the evaluation of the mischief for the rule measurements by missing values is demonstrated.