Fast rare itemset mining in uncertain database

Abstract

Rare itemset mining in uncertain database finds itemsets which have their support no less than a given probability minimum threshold. Discovering probabilistic rare itemset from a uncertain database can help user in better decision making by focusing on information that it is important but rare to something. Usually, frequent itemset and its relevant researches have gotten a lot of attention in the previous works, and much more efficient and effective algorithms was proposed to obtain valued knowledge under various constrained scenarios. Compared with frequent itemset mining, especially probabilistic rare itemset mining are rarely studied. Therefore, we present the concept of mining rare itemsets in uncertain data based on rare probabilities, and propose two algorithms, i.e. the bottom-up algorithm and the top-down algorithm. On this basis, two pruning strategies are designed to reduce the computational cost by traversing from long itemsets to short itemsets. Finally, we validate and evaluate the correctness and effectiveness of the algorithms in several experiments.