Mining Perfectly Sporadic Rules with Two Thresholds

A sporadic rule is an association rule which has low support but high confidence. It is divided into two types: perfectly and imperfectly sporadic rules. In this paper, we describe an efficient algorithm to mine perfectly sporadic rules by proposing a problem of mining perfectly sporadic rules with two thresholds and developing a MCPSI (mining closed perfectly sporadic itemsets) algorithm to find perfectly sporadic itemsets with two thresholds. Unlike the previous approaches, the development of MCPSI algorithm is based on the pruning of the closed itemset lattice, therefore efficiency of the algorithm can be improved via reducing a search space and removing redundant imperfectly sporadic rules with two thresholds. Experiments comparing MCPSI to Apriori-Inverse on the same databases also proved this conclusion.