A fast forward algorithm on discovery large itemsets

Discovering large itemsets is the key problem of algorithm for data mining. In this paper, the support vector of itemsets is presented. The algorithm LIG can prognosticate the capability of a large k-itemset extending a candidate k+1-itemset by calculating support vector, so the size of candidate set has been reduced and the efficiency of algorithm has been raised. As the number of items is very large, the size of candidate set is huge. Main memory can´t load the entire candidate set. For reduce I/O swap, the algorithm builds the candidate hash tree based on the estimated support of candidate itemsets. The performance of algorithm is better.