Frequent pattern network mining algorithm based on transaction-item association matrix

To increase the efficiency of data mining is the emphasis in this field at present. Aiming at the difficulties of data maintaining and updating in association rule mining FP-growth algorithm, this paper proposes a FP-network model which compresses the data needed in association rule mining in a FP-network. Compared with the primary FP-tree model, FP-network is undirected, which enlarge the scale of transaction storage; furthermore, the FP-network is stored through the definition of transaction-item association matrix, it is convenient to make association rule mining on the basic of defining node capability. Experiment results show that the FP-network mining association rule algorithm proposed by this letter not only inherits the merits of FP-growth algorithm, but also maintains and updates data conveniently. It improves the efficiency of association rule mining.