Optimal usage of pessimistic association rules in cost effective decision making

Abstract

Strength of an association rule is expressed in terms of interest between antecedent and consequent which is measured by using objective interestingness measures. The rules that contain at least user defined minimum interestingness are treated as Strong Association Rules (SAR). All of the SARs, extracted from the same database have no equal importance in knowledge discovery. As the user has limited resource to fix up the minimum interestingness, so there is a chance of generation of some SARs with high dissociation. Conceptually dissociation is opposite of association. As a matter of fact this type of SARs has pessimistic significance in knowledge discovery. It seems that some of the SARs are not really strong rules. This paper tries to find out such type of pseudo SARs termed as Pessimistic Association Rules (PAR). Dealing with PARs may not satisfy the user's expectation in decision making. Thus rejection of PARs may lead to a cost effective way to get a concrete decision. Experimental analysis proves the effectiveness of the proposed concept.