A GLOBAL RULE PRUNING FOR ACO IN CLASSIFICATION TASK OF DATA MINING

Abstract

Ant colony optimization algorithms have been used to extract classification rules in data mining. Different variants of ACO have been proposed and successfully applied to discover a list of classification rules. But most of these algorithms follow sequential covering of the algorithm where a rule is pruned without taking into consideration its effects on the whole list of the rules i.e. each rule is pruned individually. Moreover, the relation between the different terms in the rule generated is not taken into consideration. We have proposed a new rule pruning procedure that takes into account the association between the terms in a generated rule during pruning.