An incremental algorithm for frequent pattern mining based on bit-sequence

Abstract

In real-word applications, the target data is changed with time in association rules mining, and then existed association rules will also be changed, so the incremental mining algorithm should be developed. In this paper, we present IFPM-BS, a new incremental approach for mining frequent pattern. In this algorithm, we adopt the bit-sequence to compress the database to save the memory space, and the concept of the pre-large itemset is cited, the original database is not rescanned until a number of new transactions have been inserted; And then we define the concept of depth-branch and FLUFP-tree(Fast location Updated Frequent Pattern Tree) structure, the depth-branch is contained in the Header-Table of the FLUFP-tree to locate the nodes of the tree. If a pre-large item of the original database is changed into the large item, we can locate the nodes of the tree fast according to the matrix structure and the Header-Table information, the projected bit-sequences are contained in the matrix, and we don’t need to determine which transactions contain the item in the original database; Finally we can get the mining result by FP-growth algorithm. Experimental results also show that the proposed IFPM-BS algorithm can reduce the cost of time and improve mining efficiency.