Vertical Mining of Frequent Patterns Using Diffset Groups

Frequent patterns discovery is a core functionality used in many mining tasks and large broad application. In this paper, we present a new algorithm, VMUDG, for vertical mining of frequent itemsets. The proposed algorithm adapts a new efficient approach that classifies all frequent 2-itemsets into separate groups according to their diffsets. Using these groups, the proposed algorithm offers three new distinct features; First, it allows calculating the support of N itemsets (N is > 0) using one calculation process rather than N calculation processes. Second, it offers a chance to reduce the time needed for the manipulation of the itemsets diffsets. Third, it minimizes the need for checking the frequency condition for every itemset. A performance study of the proposed algorithm has been conducted. Several experiments show that the algorithm outperforms the well known dEclat algorithm.