Non-Redundant Rare Itemset Generation

Rare itemsets are likely to be of great interest because they often relate to high-impact transactions which may give rise to rules of great practical significance. Research into the rare association rule mining problem has gained momentum in the recent past. In this paper, we propose a novel approach that captures such rare rules while ensuring that redundant rules are eliminated. Extensive testing on real-world datasets from the UCI repository confirm that our approach outperforms both the Apriori-Inverse(Koh et al. 2006) and Relative Support (Yun et al. 2003) algorithms.