Strongest Association Rules Mining for Personalized Recommendation

Abstract

The article proposed the notion of strongest association rules (SAR), developed a matrix-based algorithm for mining SAR set. As the subset of the whole association rule set, SAR set includes much less rules with the special suitable form for personalized recommendation without information loss. With the SAR set mining algorithm, the transaction database is only scanned for once, the matrix scale becomes smaller and smaller, so that the mining efficiency is improved. Experiments with three data sets show that the number of rules in SAR set in average is only 26.2 percent of the total number of whole association rules, which mitigates the explosion of association rules.