Itemsets with relatively low support values are important since they usually suggest highly confident association rules, which are useful in applications such as recommendation systems and medical data analysis. However, most existing algorithms are mainly designed to mine frequent patterns and thus are time consuming in generating low support patterns. There are also a few algorithms focus on low support patterns but not efficient enough. Therefore, we propose here a low support closed pattern mining algorithm, utilizing top-down lattice traversing and novel closeness checking/pruning techniques. Extensive experiments show that our method is much more efficient to mine low support closed patterns than available alternatives.
展开▼
