Profile association rule mining using tests of hypotheses without support threshold

Abstract

Association rule mining has been a core research topic in data mining. Most of the past researches focused on discovering relationships among items in the transaction database. In addition, mining algorithms for discovering association rules need the support threshold to discover frequent itemsets. But the essence of association rule mining is to find very associated relationships among itemsets not to discover frequent itemsets. In this paper, we deal with mining the relationships among the customer profile information and the purchased items. We make the sample databases from the original database and use the tests of hypotheses on the interestingness of the rules from the sample data. Our approach can speed up mining process by storing the sample database into main memory and provide insights by presenting the rules of low support but high association. © Springer-Verlag Berlin Heidelberg 2005.