A Fast Bidirectional Method for Mining Maximal Frequent Itemsets

In this paper, a fast bi-directional method and an efficient data compression method for mining maximal frequent itemsets is proposed. A flexible search method is given, which exploits the advantages of bottom-up and up-bottom strategies. The compression technique use the Prime number characteristics to transform transaction data into a positive integer and can efficiently reduce the size of transaction database. This method can mine maximal frequent itemsets according to different user-defined minimum support with only one scan of original database. Theoretical and experimental analysis shows that the proposed method is scalable and efficient for mining maximal frequent itemsets.