ojme.01.2017.01.03

In order to mining frequent patterns more efficiently, in this paper we propose a matrix mining method based on the FP-tree to generate the frequent patterns. This method overcomes the recursive generation of conditional pattern trees which cost the vast majority time and memories of the FP-tree grow method. It mines the frequent itemsets with a local frequent items matrix based on the item in question which needs twice scans of the FP-tree to construct the matrix, then the frequent patterns mining will be done through the operation on the matrix. We provide an implementation of the method with a simple example and proved the efficiency of the proposed method comparing with conditional pattern tree by discussing the structure and principle of the method.