Frequent pattern generation in association rule mining using weighted support

Abstract

Determination of frequent sets from a large database is the key to Association Rule mining from the point view of efficiency of algorithms to scale up and discovering frequent sets which lead to useful association rules. Some of the existing methods have either very low or very high pruning, which is the cause of generation of larger or lesser number of frequent patterns. In this paper we have adopted a balanced approach for frequent pattern selection. Our proposed measure weighted support considers association and dissociation among items as well as the impact of null transactions on them for frequent set generation. Impact of increasing itemset size on weighted support gives rise to variable threshold The experimental results obtained after implementation of the proposed algorithm justify the approach.