Probabilistic maximal frequent itemset mining over uncertain databases

Abstract

We focus on the problem of mining probabilistic maximal frequent itemsets. In this paper, we define the probabilistic maximal frequent itemset, which provides a better view on how to obtain the pruning strategies. In terms of the concept, a tree-based index PMFIT is constructed to record the probabilistic frequent itemsets. Then, a depth first algorithm PMFIM is proposed to bottom-up generate the results, in which the support and expected support are used to estimate the range of probabilistic support, which can infer the frequency of an itemset with much less runtime and memory usage; in addition, the superset pruning is employed to further reduce the mining cost. Theoretical analysis and experimental studies demonstrate that our proposed algorithm spends less computing time and memory, and significantly outperforms the TODIS-MAX[20] state-of-the-art algorithm.