A Matrix-Based Approach for Frequent Itemsets Mining

Recent advances in computer technology in terms of speed, cost, tremendous amount of computing power and decreased data processing time has spurred increased interest in data mining applications to extract useful knowledge from data. Discovering association rules that identify relationships among sets of items is an important problem in data mining. Finding frequent itemsets is computationally the most expensive step in association rule discovery and therefore it has attracted significant research attention. In this paper, a novel approach for mining complete frequent itemsets is presented. The algorithm is partially based on FP-tree hypothesis which generates a candidate set of large 2-itemsets, a matrix is formed using the support of 2-itemsets, as a result generating all possible frequent k-itemsets in the database.